Aging-related Deficiency of CD28 Expression in CD4 1 T Cells Is Associated with the Loss of Gene-specific Nuclear Factor Binding Activity*

Changes in T cell populations and concomitant per-turbation of T cell effector functions have been postulated to account for many aging-related immune dysfunctions. Here, we report that high frequencies of CD28 null CD4 1 T cells were found in elderly individuals. Because deviations in the function of these unusual CD4 1 T cells might be directly related to CD28 deficiency, we examined the molecular basis for the loss of CD28 expression in CD4 1 T cells. In reporter gene bioassays, the minimal promoter of the CD28 gene was mapped to the proximal 400 base pairs (bp) of the 5 * untranslated region. CD28 deficiency was associated with the loss of two noncompeting binding activities within a 67-bp segment of the minimal promoter. These binding activities were not competed by consensus Ets , Elk , or AP3 motifs that were found within the sequence stretch. The DNA-protein complexes were also not recognized by antibodies to Ets-related transcription factors. Furthermore, introduction of mutations into the 67-bp segment at positions corresponding to the two DNA-protein interaction sites, i.e. nucleotides spanning 2 206 to 2 179 and 2 171 to 2 148, resulted in the loss of specific nuclear factor binding activities and the abro-gation of promoter activity. These observations impli-cate at least two regulatory motifs in the constitutive expression of CD28. The loss of binding activity of trans acting factors specific for these sequences may contribute to the accumulation CD4 1 CD28 null T cells during aging. their of CD4, CD28 cytometry. Cell lines used in derived selected individuals percentages ( $ 20%) of CD28 null CD4 1 T cells. CD3 CD4 cells from The CD28 null and the CD28 sublines two CD28 cytometry. parent cell CD28 null sequence with the associated SV40 site is located downstream of the minimal SV40 This plasmid as one of the control plasmids in transient transfection assays, as well as the cloning vector for the construction of a series of hGFP reporter plasmids containing fragments of the CD28 5 9 -UTR. Truncation variants of the 5 9 -UTR were by PCR and cloned into the II and Hin dIII sites of pSVGFPAmpR after excision of the minimal SV40 promoter. plasmids were amplified in bac-teria as described authenticated chloride density then EMSAs either as binding probes or as competitors to the wild type sequence. Preparation of nuclear extracts, conditions of binding reactions, and resolution and visualization of gel shifts were as described above. Sequence Mapping of the CD28 5 9 -UTR— The cloned 600-bp segment of the CD28 5 9 -UTR was subjected to transcription factor mapping using the Map program in the GCG software interfaced with the transcription factor data base. Mapping algorithms were set at high strin- gency in that the relevant nuclear factor binding sites were identified with exact matches or with a single allowable mismatch.

dence that changes in T cell populations and concomitant changes in T cell effector functions may underlie many agingrelated immune dysfunctions (1). Because of thymic involution from the onset of reproductive maturity, aging leads to the inability to generate new T cells. Thus, there is a progressive replacement of naive (CD45RAϩ) T cells with those having a memory (CD45ROϩ, CD29 hi , and CD44 hi ) phenotype (2)(3)(4)(5). Paradoxically, expansion of the memory T cell population is not associated with enhanced memory immune responses in the elderly. Although the reason for this phenomenon is not known, previous studies indicate that T cells undergo functional changes with aging. Defects in signal transduction and, consequently, decreased proliferation and interleukin (IL) 1 -2 production have been reported (6 -11). Other reported aging-related changes in T cell function are the inability of T cells to support B cell proliferation and immunoglobulin production (1,12). Thus, T cell-dependent B cell responses are impaired with aging as demonstrated by the marked reduction in the efficacy of vaccines in the elderly (13,14).
In an attempt to identify mechanisms of impaired immune responses, we studied a cohort of elderly individuals. Phenotypic T cell profiles of these individuals indicate that a subgroup had accumulated CD4ϩ T cells deficient in the expression of CD28, a membrane glycoprotein typically found on CD4ϩ T cells that provides the requisite costimulatory signal for T cell activation (15,16). In the elderly, these CD28 null cells can comprise up to 45% of the total CD4ϩ T cell compartment. A similar CD4ϩ T cell subset is typically found among patients with rheumatoid arthritis. Analysis of the T cell receptor usage of these in vivo expanded CD4ϩCD28 null T cells (17,18) indicated that they are oligoclonal (19) and persist for many years (20).
We propose that the emergence of CD4ϩCD28 null T cells may reflect an aging immune system. Consistent with this hypothesis are the features of these cells, namely, their in vivo clonality and longevity, memory phenotype, and unusual functional profiles, including defective B cell help activity (17)(18)(19)(20)(21)(22). All of these characteristics are considered to be hallmarks of an aging immune system (1). Inasmuch as CD4ϩCD28 null T cells appear to be a distinct subset of functional T cells in the elderly but not among younger individuals, further assessment of the immunobiology of these cells may provide insight into the cellular processes involved in immunosenescence.
Our interest in studying the biology of CD4ϩCD28 null T cells in humans lies in the well documented role of CD28 in immune responses as the major costimulatory molecule required for T cell activation (15,16). In the mouse, the CD28 molecule is expressed on virtually all CD4ϩ T cells and has been unequivocally shown to be required for the maintenance of T cell proliferation, IL-2 production, and the de novo synthesis of many T cell-specific genes. In the absence of CD28-mediated signals, T cell recognition of antigen results in anergy or the induction of apoptosis. Perhaps the most compelling evidence of the critical requirement for CD28 in T cell-mediated immunity is the CD28-deficient mouse, which exhibits various forms of immune dysfunctions (23)(24)(25)(26).
Among the important issues related to the biology and emergence of the CD4ϩCD28 null T cells is the molecular basis for the deficiency of CD28 expression. The genomic and cDNA structures of the gene have been reported previously (27,28), but the structural and biochemical requirements for gene transcription remain to be elucidated. Therefore, we initiated studies to examine the structural requirements for the constitutive expression of CD28. It is important to note that although CD28 is constitutively expressed on CD4ϩ T cells, previous studies showed that the levels of its expression are not static. The level of CD28 expression has been shown to either transiently increase or decrease, depending on the stimulus (29,30), implying that the levels of its expression profoundly influence its costimulatory function.
The observed deficiencies in cell surface expression of CD28 among CD4ϩ T cells is correlated with a lack of CD28 mRNA. 2,3 Similar observations have also been reported for transformed CD4ϩCD28 null T cells without apparent genomic lesions of CD28 itself (27). Therefore, we evaluated the hypothesis that CD28 deficiency in human CD4ϩ T cells is associated with a transcriptional block. In the present paper, bioassays were carried out to assess the transcriptional activity of the CD28 gene promoter in CD28-deficient cells. Additionally, studies were conducted to identify sequence motifs that may be associated with the differential surface expression of CD28 in primary human CD4ϩ T cells. Given the important role of CD28 in the immune response, elucidation of the biochemical and molecular basis for the modulation of its expression will facilitate our understanding of the biology of CD4ϩCD28 null T cells. Because these unusual T cells appear to be a feature of the elderly, studies of the pathways controlling CD28 expression will provide insight on molecular mechanisms associated with the emergence of CD28 null cells in the aging CD4 T cell compartment.

MATERIALS AND METHODS
Study Population-A total of 60 Caucasians who did not have a personal or family history of autoimmune diseases were chosen. Mononuclear cells were isolated from heparinized peripheral blood by isopycnic centrifugation over Ficoll-Hypaque (Amersham Pharmacia Biotech). Cells were subjected to standard immunofluorescence staining with monoclonal antibodies to CD3, CD4, and CD28 (Becton Dickinson Immunocytometry Systems, San Jose, CA). Cells were fixed with 1% paraformaldehyde in phosphate-buffered saline, and the lymphocyte population was analyzed by flow cytometry using either a FACScan or FACSVantage cytometer (Becton Dickinson Immunocytometry Systems). All analyses of the staining profiles of lymphocytes were carried out using PC-LYSIS II software (Becton Dickinson Immunocytometry Systems).
The frequency patterns of CD28ϩ and CD28 null cells among CD3ϩCD4ϩ lymphocytes were analyzed as a function of age. Statistical analyses of the frequencies were performed using the Mann-Whitney test (SYSTAT for Windows, SYSTAT, Inc., Evanston, IL).
Cell lines used in the study were derived from selected individuals with high percentages (Ն20%) of CD28 null CD4ϩ T cells. CD3ϩCD4ϩ cells were isolated from peripheral lymphocytes by standard fluorescence-activated cell sorting procedures and propagated in vitro. The CD28 null and the CD28ϩ sublines were generated from their respective parental CD4ϩ lines by at least two successive cell sorting procedures. Purity of the sublines for their CD28 phenotype was assessed periodically by flow cytometry. With these approaches, the parent cell lines maintained to date are 30 -40% CD4ϩCD28 null , whereas the CD28ϩ sublines are generally Ն98% pure, and the CD28 null sublines are 100% CD28-negative.
Jurkat and HUT78 cells were obtained from ATCC (Rockville, MD). Both cell lines were maintained in antibiotic-free RPMI 1640 medium supplemented with 2 mM L-glutamine. Jurkat cells were grown in 5% fetal calf serum (Summit Biotech) and 5% newborn calf serum (Life Technologies), and HUT78 cells were grown in 10% fetal calf serum. Expansion of HUT78 was enhanced by the addition of recombinant human IL-2 to a final concentration of 5 units/ml. Both cell lines were also maintained at a cell density of 10 7 cells/ml while in culture. Cells were cultured in a humidified 37°C/5% CO 2 incubator. Under these culture conditions, high efficiencies of transfection (see below) have been achieved.
Isolation of the CD28 5Ј-Untranslated Region (UTR)-The genomic template for the cloning of the CD28 gene 5Ј-UTR was obtained by a combined strategy (see Fig. 2) of restriction enzyme digestion, Southern blotting, and by polymerase chain reaction (PCR). Genomic DNA isolated from peripheral blood lymphocytes of a healthy donor was prepared using standard techniques and digested with a mixture of restriction enzymes containing HindIII, BglII, and ScaI. Restriction fragments were fractionated by agarose gel electrophoresis and blotted onto nylon membranes. The appropriate fragment was identified by hybridization with a synthetic 41-bp oligonucleotide probe (see Fig. 2B, probe 41) corresponding to a segment within the CD28 5Ј-UTR described previously (28). This 41-bp segment of the 5Ј-UTR contains the putative Ets1 and AP3 binding motifs. This combination of the two motifs within the 41-bp segment was found to be unique for the CD28 gene. Ets and AP3 motifs were determined by structural analysis of the published sequence of the CD28 5Ј-UTR (28) using the Map program in GCG software (Genetics Computer Group, University of Wisconsin, Madison, WI) interfaced with the transcription factor data base (NCBI, National Institutes of Health, Bethesda, MD).
The appropriate restriction fragment was purified and used as a template for the isolation of the 5Ј-UTR by PCR with gene-specific primers using the Expand PCR System (Boehringer Mannheim). All PCR primers used in this study contained an engineered 5Ј restriction enzyme site to facilitate subsequent cloning into plasmid vectors. Synthesis and purification of oligonucleotide probes and primers and the PCR conditions have been described (31).
The appropriate PCR amplification products were identified by the Southern hybridization procedure described above. These PCR products were digested with the appropriate restriction enzyme, purified by GeneClean (BIO 101, La Jolla, CA), and cloned into a cloning/expression plasmid (see below). Plasmids were introduced into Escherichia coli DH5␣ (Life Technologies, Inc.) by standard transformation and drug selection protocols. Recombinants were screened initially by colony lift hybridization using the 41-bp probe followed by the direct PCR of positively hybridizing colonies using the same amplification primers used in the initial PCR. Clones containing the CD28 5Ј-UTR were authenticated by sequencing.
Reporter Gene Bioassays-For the present studies, a cloning/reporter plasmid was specifically created. The plasmid, pSVGFPAmpR, was a fusion of the appropriate BamHI-HindIII fragments from pGLPromoter (Promega, Madison, WI) and phGFP-S65T (CLONTECH). It contains the ampicillin resistance gene, f1ORI, a multiple cloning site, and the minimal SV40 promoter. The humanized green fluorescence protein (hGFP) cDNA sequence with the associated SV40 poly(A) site is located downstream of the minimal SV40 promoter. This plasmid was used as one of the control plasmids in transient transfection assays, as well as the cloning vector for the construction of a series of hGFP reporter plasmids containing fragments of the CD28 5Ј-UTR.
Truncation variants of the CD28 5Ј-UTR were generated by PCR and cloned into the BglII and HindIII sites of pSVGFPAmpR after excision of the minimal SV40 promoter. These plasmids were amplified in bacteria as described above. Sequencing authenticated all CD28 5Ј UTR reporters generated, and two independent clones for each truncation variant were chosen for subsequent bioassays. All plasmids used in reporter bioassays were purified by cesium chloride density centrifugation. The plasmid preparations were then dialyzed against 1ϫ Tris-EDTA, desalted by column chromatography (NAP-10 columns, Amersham Pharmacia Biotech), and stored at Ϫ20°C.
Reporter plasmids were transiently transfected into Jurkat and HUT78 cells by electroporation. 1 ϫ 10 7 cells resuspended in 300 l of drug-free culture medium were aliquoted to electroporation cuvettes (4-mm gap Cuvette Plus, BTX, San Diego, CA). 28 g of each hGFP reporter, resuspended in 50 l of serum-free drug-free RPMI medium, was added to the cell suspension. The cell-DNA mixture was incubated on ice for 10 min and then electrically pulsed at 350 V with a capacitance of 960 F and resistance of 129 ohms using an electroporator (ElectroCell 600, BTX Electroporation Systems). Cuvettes were immediately transferred onto ice and incubated for 10 min. About 1 ml of culture medium was then added to the cuvettes to resuspend the cells and subsequently transferred to sterile 6-ml tubes. Culture medium was added to the tubes to a final volume of 3.5 ml, and the cells were incubated overnight in a tissue culture incubator. hGFP expression was determined by flow cytometry 20 h after transfection. The analyses were carried out by electronically gating on viable cells by standard propidium iodide staining. For each experiment, a total of 50,000 events were collected for quantitative analysis.
In experiments involving mitogens, transiently transfected cells were initially incubated in tissue culture medium for 4 h before the mitogens were added. Phorbol 12-myristyl 13-acetate (PMA) (Sigma Chemical Co.) was added to a final concentration of 50 ng/ml, and ionomycin (Sigma) was added to a final concentration of 10 nM. In initial experiments (data not shown), these concentrations of PMA and ionomycin were found to promote the proliferation of primary CD4ϩ T cells and IL-2 production by Jurkat cells under the culture conditions described above.
For each transfection experiment, two control transfections were carried out. One of the controls involved the transfection of pSVGF-PAmpR that served as the basal/vector control. The other involved transfection of phGFP-S65T (CLONTECH), a hGFP reporter under the control of the CMV promoter. hGFP reporter gene bioassays conducted also involved the cotransfection of 2 g of the luciferase plasmid pCMVLuc (32). This was done to normalize for transfection efficiency of each test hGFP reporter. Luciferase activities were determined by bioluminescence using a bioassay kit (Promega), and photoemissions were measured by a luminometer (Lumat LB9501, Berthold Analytical, Nashua, NH).
Electrophoretic Mobility Shift Assays (EMSAs)-In the present work, EMSAs were conducted using nuclear extracts from both CD28ϩ and CD28 null CD4ϩ T cell lines and clones described above. Additional extracts from Jurkat, HUT78, other non-T cell ATCC lines (HeLa, K562, RD, and U937), and the Epstein-Barr virus-transformed B lymphoblastoid cell line HT10 were also prepared. Binding probes used in these assays were as indicated (see "Results" and Fig. 4). In competitive EMSAs, competitors were added to the binding reaction at 5-300-fold excess concentrations relative to the binding probe. The conditions for the preparation of nuclear extracts, radiolabeling of oligonucleotide probes, binding reactions, nondenaturing polyacrylamide electrophoresis, and visualization of gel shifts have been described (33).
Mutational Analysis-Results of EMSAs indicated that there are two sequence motifs in the CD28 promoter 5Ј-UTR that bound nuclear factors from CD28ϩ, but not CD28 null , CD4ϩ T cells (see "Results"). To further evaluate the role of these two motifs in the promoter activity of the 5Ј-UTR, site-specific mutations were introduced into reporter constructs containing the minimal CD28 promoter (see "Results" and Fig.  3). Cluster of overlapping mutations, as indicated, were introduced by the gene splicing by overlap extension technique described previously (31). For these studies, the mutants were cloned into the BglII-HindIII sites of the vector, pGLPromoter (Promega) after the excision of the minimal SV40 promoter. Two independent clones for each mutant were selected. The plasmids were purified by cesium chloride density centrifugation and used in transient transfection assays with Jurkat cells as described above. For each transfection experiment, the plasmid pCMVLuc (32) was used as positive control, and the pGLPromoter plasmid served as vector control. Transfection efficiency was normalized by the cotransfection of 2 g of the ␤-galactosidase reporter, pCMV␤gal (32). Luciferase and ␤-galactosidase activities were determined by bioluminescence using the Dual-Light Assay kit (Promega), and photoemissions were measured by a luminometer (Lumat LB9501).
Mutants that showed ablation of reporter activity were identified, and their corresponding sequences were synthesized. Such oligonucleotides were used in EMSAs either as binding probes or as competitors to the wild type sequence. Preparation of nuclear extracts, conditions of binding reactions, and resolution and visualization of gel shifts were as described above.
Sequence Mapping of the CD28 5Ј-UTR-The cloned 600-bp segment of the CD28 5Ј-UTR was subjected to transcription factor mapping using the Map program in the GCG software interfaced with the transcription factor data base. Mapping algorithms were set at high stringency in that the relevant nuclear factor binding sites were identified with exact matches or with a single allowable mismatch.

Deficiency of CD28 Expression in Human CD4ϩ T Cells Is
Correlated with Age-Our previous studies indicated that CD4ϩCD28 null T cells are extremely infrequent in the majority of healthy individuals; however, a subset of individuals showed increased frequencies (18). We examined an extended cohort of healthy Caucasian individuals to identify factors associated with the increased numbers of CD4ϩ CD28 null cells. As shown in Fig. 1, the frequency of CD4ϩ CD28 null T cells was significantly correlated with age. These cells were rarely found among individuals younger than 40 years. In contrast, many of the elderly have a marked increase in the frequency of CD4ϩCD28 null T cells. In some elderly individuals, these cells comprised up to 45% of the total CD4ϩ T cells (Fig. 1, inset). It is important to note that there were also elderly individuals who had extremely low frequencies of these unusual cells (Ͻ1% of the total CD4ϩ T cells) and those who completely lacked them. Conversely, there were occasionally those younger than 40 years with high frequencies (Ͼ2%) of CD4ϩCD28 null T cells. These results indicate that there are other age-independent factors influencing the presence of these cells in the peripheral blood.
The Promoter Activity of the CD28 5Ј-UTR Is Localized to the Proximal 400-bp Region-The deficiency in the cell surface expression of CD28 in CD4ϩ T cells was previously shown to be due to the uniform lack of CD28 mRNA. 2, 3 By standard reverse transcription-PCR techniques, none of the splice variants of CD28 mRNA (27,28) were detectable. To test the hypothesis that a transcriptional block might account for the deficiency in CD28 expression, studies were initiated to isolate the 5Ј-UTR of the gene for functional assays. The genomic structure of the human CD28 gene has been described (28). The restriction map of the gene has also been reconstructed by the mapping of contiguous phage clones. However, the sequence encompassing the entire gene domain remains to be determined. The regions of the gene that have been sequenced include 760 bp of the 5Ј-UTR, exons 1-4, and parts of introns 1-3. Intronic sequences that have been determined are those that immediately flank the exons. Using this information, we conducted a limited restriction enzyme analysis to isolate the appropriate genomic template for the cloning of the CD28 5Ј-UTR. For these studies, HindIII, BglII, and ScaI enzyme digestion was performed. As indicated previously (28), ScaI and HindIII sites are located within introns 1 and 2, respectively ( Fig. 2A). Thus, a mixture of these enzymes was predicted to yield restriction fragments containing the CD28 5Ј-UTR. By Southern hybridization using a 41-bp sequence (Fig. 2B), the 5Ј UTR was found to be localized within a Ϸ2.5 kb HindIII fragment ( Fig. 2A). This fragment was within larger BglII and ScaI fragments because digestion of DNA with the all three enzymes or a combination of HindIII with either BglII or ScaI yielded hybridizing fragments of the same size as the HindIII digested DNA. These results indicated that ScaI, BglII, and HindIII sites are situated on either side of exon 1. Using the Map programs in the GCG software, none of these restriction sites were found within exon 1 or the 760-bp sequence of the 5Ј-UTR (data not shown).
Based on the results of the Southern blotting experiments, the genomic template for the cloning of the 5Ј-UTR by PCR was isolated. Using gene-specific primers, the proximal 600-bp of the 5Ј-UTR have been cloned. Because mechanisms involved in the normal expression of CD28 are not yet known, the cloned proximal region of the 5Ј-UTR was subsequently used in reporter gene bioassays to examine its promoter activity in transiently transfected cells. The cloned 600-bp region of the 5Ј-UTR and its truncation variants were cloned upstream of the hGFP cDNA sequence to generate a panel of reporter plasmids (Fig. 3A). These were transiently transfected into Jurkat cells, which express high levels of CD28 (data not shown), and hGFP expression was subsequently measured. Results of experiments showed that only cells transfected with reporters containing the proximal 400-bp segment of the 5Ј-UTR (plasmid 42) had significant hGFP activities (Fig. 3B). Such reporter activities were comparable to those seen with cells transfected with hGFP reporters under the control of the CMV promoter. Cells transfected with reporters containing the full 600-bp 5Ј-UTR (plasmid 52) showed low to negligible reporter activities, equivalent to those seen with cells transfected with the reporters containing the minimal SV40 promoter, which lacks enhancer activity.
Previous studies indicated that although CD28 is constitutively expressed, the surface levels of CD28 expression may be modulated by mitogens (30). Thus, hGFP reporter bioassays were conducted to examine whether such mitogen-induced modulation may occur at the level of the gene promoter. As depicted in Fig. 3C, addition of 50 ng/ml PMA and 10 nM ionomycin to transiently transfected cells elicited reductions in hGFP levels. These reductions were evident in cells transfected with reporters containing the proximal 400 bp of the 5Ј-UTR (plasmid 42). Addition of mitogens to cells transfected with the other reporter constructs did not affect hGFP levels. The hGFP activities of cells transfected with CMV-driven reporters were also not significantly affected by incubation with mitogens.
The Promoter Activity of the Proximal 400-bp Segment of the 5Ј-UTR Is Suppressed in CD28 null T Cells-As indicated earlier, the deficiency of CD28 expression among CD4ϩ T cells was associated with a uniform lack of CD28 mRNA. 2,3 This suggested that a transcriptional block may be responsible for the lack of cell surface expression. To test this hypothesis, reporter gene bioassays were conducted using the CD4ϩCD28 null T cell line HUT78. As we have found among our T cell clones, the lack of CD28 expression in HUT78 also occurred at the mRNA level without any apparent lesions in the genomic sequence of the gene (27). Results of transient transfection experiments with hGFP reporters showed that the promoter activity of the 400-bp segment of the CD28 5Ј-UTR was suppressed in HUT78 cells (Fig. 3D). The hGFP activities of cells transfected with the CMV promoter-driven reporters were maintained as expected. The negligible hGFP activities of HUT78 cells transfected with the other reporters were not significantly different from similarly transfected Jurkat cells.
A Lack of DNA-Protein Complex Formation with a 67-bp Sequence of the 5Ј-UTR Is Associated with a CD28-deficient Phenotype-The observation that the promoter activity of proximal 400 bp of the CD28 5Ј-UTR was suppressed in HUT78 suggested that deficiency in CD28 expression may be associated with the lack of a relevant trans-acting factor(s). To eval- uate this hypothesis, experiments were conducted to determine whether segments of the 400-bp region exhibited differential in vitro binding activities for nuclear factors from untransformed CD28ϩ and CD28 null CD4ϩ T cells. In initial EMSAs, a 41-bp sequence (referred to as probe 41) was used (Figs. 2B and 4). Results of binding assays showed differential binding activities of probe 41 (Fig. 5). Binding activities to the probe were consistently observed with extracts from untransformed CD4ϩCD28ϩ T cell lines or clones and from Jurkat cells. In contrast, extracts from CD4ϩCD28 null T cells, either transformed or untransformed, showed negligible binding activities. This differential binding activity was shown to localize to the opposite ends of probe 41. EMSAs using segments of probe 41 (Fig. 4), namely, the 5Ј half (probe 41A) and the 3Ј half (probe 41B), also showed differential binding activities (factors A and B) that correlated with the CD28 phenotype (Fig. 5). In contrast, binding assays using a probe corresponding to the central segment (probe 41C) did not show any binding activity regardless of the extract used in the assay. Interestingly, the binding activities seen with the 3Ј half (probe 41B) showed DNA-protein complexes with gel mobilities that were very similar to those seen with the intact 41-bp sequence (probe 41, factor A). This is in marked contrast to the activities seen with the 5Ј half (probe 41, factor B) which showed faster gel mobilities. Binding assays involving the same nuclear extracts using probes for the transcription factor SP1 (33) showed little difference among the T cell lines and clones examined. Moreover, EMSAs conducted using extracts from non-T cell lines also showed variable binding profiles (data not shown) that were unlike those seen with T cells as depicted in Fig. 5. Among the non-T cells examined were HeLa (epithelioid carcinoma), K562 (erythroid cell line), U937 (promonocytic line), RD (rhabdomyosarcoma), and HT10 (Epstein-Barr virus-transformed B lymphoblastoid line). These observations indicated that the differential binding activities of the 41-bp sequence or its specific segments were clearly correlated with the CD28 phenotype of the cells.
The differences between the gel mobilities of DNA-protein complexes formed with probes 41A and 41B suggested that these were two distinct activities. Because these probes were of the same length, the observed binding activities likely represent two different DNA-binding protein complexes. However, this is confounded by the observation that DNA-protein complexes formed with the intact 41-bp sequence showed very similar mobilities with that of the 3Ј half (probe 41B). These suggested that there are two nonoverlapping sites localized to either side of the sequence. To confirm these findings, probes including nucleotide positions upstream of the 41-bp probe were used (Fig. 4). Experiments showed that the binding activity seen with the 5Ј half (probe 41A) extended to nucleotide positions flanking the 41-bp sequence, i.e. probes 4441A and 44 (Fig. 5). DNA-protein complexes seen with these probes showed gel mobilities that were very similar to those with the 5Ј half (probe 41A). Similarly, the binding activities of the 3Ј half (probe 41B) were indistinguishable from probes that extended to downstream immediately flanking nucleotides (probe 5152; data not shown). Thus, the 5Ј and 3Ј binding activities were indeed distinct from each other. These observations also indicated that the segment of the CD28 5Ј-UTR that shows differential binding activities for nuclear extracts from CD28ϩ and CD28 null CD4ϩ T cells encompasses a 67-bp sequence. Incidentally, this 67-bp sequence overlaps with the presumed transcription initiation site previously reported by Lee et al. (28).
There Are Two Noncompeting DNA-Protein Interaction Sites within the 67-bp Segment of the 5Ј-UTR-To further ascertain that the two binding activities within the 67-bp sequence defined by probes 41A and 41B were nonoverlapping, competitive EMSAs were conducted. The experiments showed that probe 41A activities were not reciprocally competed by the 41B activities (Fig. 6). The respective DNA-protein complexes seen with the 41A probe were very stable in that even a 300-fold excess of competitor probe 41B or the 41B-related probe 5152 (data not shown) did not abrogate binding activity. Conversely, 41B activities were not competed by high concentrations of probe 41A or the 41A-related probes 4441A and 44. As expected, competition assays involving related probes 41A, 4441A, and 44 effectively competed each other, just as the related probes 41B and 5152 were effective competitors ( Fig. 6 and data not shown).
Additionally, these two nonoverlapping binding activities within the CD28 promoter were not competed by other sequences containing the binding motifs for known transcription factors ( Fig. 6 and data not shown). Among the sequences examined were nuclear factor B, STAT1␣, IRF1, and SP1 (33).
As indicated in Fig. 4, the 67-bp sequence from the CD28 5Ј-UTR contained the binding motifs for the Elk1, Ets1, and AP3 transcription factors. Elk1 and Ets1 are structurally related proteins (34,35) that have been shown to play an important role in the expression of T cell-specific genes (36 -40). In contrast, AP3 and AP3-like factors are not known to mediate T cell-specific gene transcription, although they have been implicated in the differentiation of myeloid cells (41) and in the synthesis of an IgE-induced mast cell-derived cytokine (42).
To determine whether Ets-related factors contribute to the

FIG. 2. Isolation of the 5-UTR of the human CD28 gene.
Genomic DNA was subjected to restriction enzyme digestion and Southern blotting to identify the appropriate restriction fragment containing the CD28 gene domain. A, based on the Southern blots and previous data (28), a diagrammatic restriction map containing the CD28 5Ј-UTR was constructed. As shown, HindIII, BglII, and ScaI sites are situated on either side of exon 1. None of these sites have been found in exon 1 or in the 760-bp sequence of the 5Ј-UTR as determined by mapping in GCG. Also indicated in the map are the previously identified ScaI and HindIII sites, the 22-kb intron 1, and exon 2. B, using PCR and genespecific primers (indicated by arrows), the proximal 600-bp region of the CD28 5Ј-UTR has been successfully cloned. Also depicted is the location and sequence of the 41mer sequence (probe 41) used as the hybridization probe in Southern blots. Base numbering of the 5Ј-UTR begins immediately upstream from the translation start codon, as shown.

FIG. 3. Assessment of the promoter activity of the CD28 5-UTR.
A, a panel of hGFP reporters containing truncation variants of the cloned 600-bp sequence of the 5Ј-UTR was generated by standard PCR and cloning procedures. B, two clones for each hGFP reporter plasmid were randomly picked and transiently transfected into Jurkat T cells. hGFP expression was measured by flow cytometry 20 h after transfection. C, similar transfection experiments were also conducted to examine the effects of mitogens on promoter activity. Transfectants were cultured in medium alone (gray bars) or in the presence of 50 ng/ml PMA and 10 nM ionomycin (black bars). D, experiments were conducted to compare the promoter activity of the constructs in the CD28 hi Jurkat T cells and the CD28 null HUT78 T cells. In all experiments, hGFP activities were normalized against luciferase activities by cotransfection with the pCMVLuc plasmid (32). Luciferase activity was measured by bioluminescence. Data depicted are the average hGFP activities of two separate transfections for each reporter clone. hGFP activities are expressed as arbitrary units relative to the hGFP activity of the CMV-driven positive control reporter phGFP-S65T (CMV), which is shown as 100 units of maximal activity. Also shown are hGFP activities of transfections with the minimal SV40 promoter-driven reporter pSVGFPAmpR (minSV). Data shown are representative of four to six independent experiments in each data panel.

FIG. 4. Sequence of a 67-bp subsegment of the proximal 400-bp region of the CD28 5-UTR.
The sequence shown is an internal segment of the 5Ј-UTR spanning positions Ϫ220 to Ϫ153. It contains putative binding motifs for Elk1, Ets, and AP3 as identified by the Map computer program in GCG interfaced with the transcription factor data base. Also shown is the 41mer sequence (probe 41) that was originally used as the hybridization probe for the isolation of the 5Ј-UTR (Fig. 2). The individual binding probes used in the EMSAs are indicated. The 67-bp segment overlaps with the previously reported transcription initiation site (28). Base numbering of the 5Ј-UTR begins immediately upstream from the translation start codon, as shown. observed binding activities of the 67-bp sequence, supershift EMSAs were conducted. Results showed that antibodies to Ets1, Elk1, or Elf did not affect the gel mobilities of DNAprotein complexes seen with either the 41A/4441A probes or the 41B/41 probes (data not shown). Furthermore, EMSAs using probes containing previously characterized Ets-specific (36) and Elk-specific (37) sequences did not show any detectable binding activity (data not shown). These results were consistent with the earlier observation that EMSAs using probes containing the presumptive CD28 Ets site (the 41C probe) did not show any detectable binding activity (Fig. 5) regardless of the T cell-derived nuclear extract used in these studies.
To further assess whether AP3 or AP3-like factors are involved in the binding activities of the CD28 67-bp segment, reciprocal competition assays were performed with probes 41B and 5152 (Fig. 4), an AP3 consensus sequence, and an AP3-like sequence of a recently described mast cell-derived IgE-induced cytokine (42). Results showed that whereas the CD28-derived probes 41B and 5152 had significant binding activity (Figs. 5 and 6 and data not shown), neither the AP3 nor the AP3-like probes had detectable binding activities. Furthermore, these two probes did not compete with the activities of probes 41B and 5152 (data not shown).
The Two DNA-Protein Interaction Sites Are Situated Immediately Downstream from the CAAT-TATA Boxes-Mapping analysis of the cloned 600-bp segment of the CD28 5Ј-UTR showed that the 67-bp segment containing the two nonoverlapping DNA-protein interaction sites lies downstream from the putative CAAT and TATA boxes (Fig. 7). As indicated by the results of gel shift assays (Figs. 5 and 6), these two sites, herein referred to as sites ␣ and ␤, lie on opposite sides of an Ets1-like binding sequence. Site ␣ was the 5Ј site defined by probes 41A and 4441A, whereas site ␤ was the 3Ј site defined by probes 41B and 5152. As indicated, sites ␣ and ␤ overlap with Elk1and AP3-like sequences, respectively. As shown by the gel shift assays, however, neither Ets1, Elk1, nor AP3 appeared to influence the observed binding activities of sites ␣ and ␤ (see above and Figs. 5 and 6).
Along the 600-bp stretch, other sequence motifs were also found (Fig. 7). Downstream from sites ␣ and ␤, there appeared to be two nuclear factor B (NFB) sites, an interferon response element (IRE), and a cAMP-responsive element (CREB). Upstream from the CAAT box were sequences with homology to AP1, AP2, and a retinoic acid response element (RARE).
Whether or not these motifs play a role in the induction of the CD28 promoter remains to be examined. Curiously, the distal 120 bp of the cloned 600-bp region of the CD28 5Ј-UTR contained two motifs with homology to a consensus repressor element (RepE) previously found to block transcriptional activities of various gene promoters.
Mutations in the Two DNA-Protein Interaction Sites Abrogate Nuclear Factor Binding and Ablate Reporter Activity of the Minimal CD28 Promoter-To evaluate the significance of the nuclear factor binding activities of sites ␣ and ␤, a panel of reporter constructs containing mutations in these CD28 promoter motifs were generated. To identify the critical subsites for both motifs, a cluster of four nucleotide substitutions were introduced across the sequence stretch, thereby generating a series of overlapping mutants. As depicted in Fig. 8, the mutagenesis yielded motif variants that ablated reporter gene activities in that mutations in either site ␣ or site ␤ independently elicited the inactivation of the minimal CD28 promoter. Mutations in site ␣ that resulted in the loss of promoter activity were within a cluster of seven centrally located nucleotide positions, i.e. defined by mutants M3 and M4 and spanning positions Ϫ194 to Ϫ186, whereas in site ␤, the inactivating mutations clustered in four to six more distal positions, i.e. defined by mutants M9, M10, and M11 and spanning positions Ϫ161 to Ϫ154. Curiously, mutations located between the two sites (mutant M5) did not ablate reporter activities.
It is important to note that the inactivating mutations within sites ␣ and ␤ were situated at the 3Ј-flank of the overlapping Elk1 and AP3 sequences, respectively (Figs. 4, 7, and 8). These results, together with the inability of Elk1 and AP3 consensus sequences to compete specific nuclear factor binding (see above), confirm that neither factors were involved in the DNAprotein complex formation of sites ␣ and ␤.
To determine whether the ablation of reporter activities by specific mutations in either site ␣ or ␤ was correlated with the loss of DNA-protein complexes, EMSAs were carried out using the indicated mutant sequences as binding and competitor probes. As depicted in Fig. 9, none of the mutants showed significant nuclear factor binding activities, unlike their wild type counterparts. Furthermore, the mutants did not compete the wild type binding sequences even at a 300-fold excess of the concentration of the binding probe. FIG. 5. Differential binding activities of the 67-bp sequence for nuclear extracts from CD28؉ and CD28 null CD4؉ T cells. Nuclear extracts from various T cell lines and T cell clones were prepared and analyzed in EMSAs. The conditions of the binding reactions were as described previously (32). About 15 g of nuclear extracts was used in all assays. The binding probes were as shown (Fig. 4). DNA-protein complexes were resolved by nondenaturing polyacrylamide gel electrophoresis using 6% gels. The gels were dried and exposed to autoradiographic film. As a system control, EMSAs were also carried out using SP1 sequences (33). Data shown are optical images of representative autoradiograms. Binding profiles shown were reproducible in at least three experiments for each of five independent preparations of nuclear extracts.  6. Localization of two noncompeting nuclear factor binding activities within the 67-bp segment of the 5-UTR. Nuclear extracts from CD4ϩCD28ϩ T cells were prepared and used in competitive EMSAs. Conditions of the binding reactions were as in Fig. 5 with the addition of increasing concentrations (5-300-fold excess; SP1, 300-fold excess only) of the indicated competitor. Data shown were obtained from experiments with extracts from sorted CD4ϩCD28ϩ subline of the DH cell line (Fig. 5). Similar results were obtained using the CD4ϩCD28ϩ T cell clone P43 and Jurkat cells (data not shown). Binding probes and competitors used are indicated. The use of previously described Ets-like (36), Elk-like (37), and AP3-like (42) sequences as competitors did not inhibit the binding activities of the indicated CD28-derived sequences (data not shown). Data depicted were reproducible in two experiments for each of three independent nuclear extract preparations.

FIG. 7.
Transcription factor mapping of the CD28 5-UTR. The cloned 600-bp segment of the CD28 5Ј-UTR was subjected to sequence mapping for DNAprotein interaction sites using the Map programs in the GCG software interfaced to the transcription factor data base. The diagram indicates the location of sequence motifs identified by exact homology or a single allowable mismatch to known transcription factor binding sequences. Also shown is the sequence stretch containing two motifs, referred to as sites ␣ and ␤, that bound nuclear factors from extracts of CD28ϩ but not CD28 null , CD4ϩ T cells (see "Results" and Figs. 4 -6).

DISCUSSION
Studies presented here address molecular mechanism(s) underlying the deficiency in CD28 expression of CD4ϩ T cells. CD4ϩ CD28 null T cells are uncommon in healthy young indi-viduals but emerge in the elderly. In some cases, these cells comprise up to 45% of the CD4ϩ T cell compartment. Given the critical role of CD28 in the immune response (15,16,23,24), defective CD28 expression may contribute to T cell dysfunction characteristic of immunosenescence. Here we show that the lack of CD28 expression correlates with the absence of two trans-acting factor binding activities within the minimal CD28 promoter.
Accumulation of CD4ϩCD28 null T Cells in an Aging Immune System-The occurrence of CD4ϩCD28 null T cells at unusually high frequencies among the elderly (Fig. 1) raises the issue of the relevance of these cells in immunity of the elderly. As has been well documented in both mice and humans, CD28 is the critical costimulatory molecule required for T cell activation, proliferation, and effector function (16). Its absence or interruption of its interaction with its ligands results in T cell anergy or the induction of apoptosis (15,23,25). In mice, the targeted deletion of CD28 resulted in a severely immunocompromised animal (24,26). Thus, the observed accumulation of CD4ϩCD28 null T cells in vivo raises the hypothesis that CD28 deficiency leads to deviations in T cell effector function and contributes to age-related immune dysfunctions.
As demonstrated previously, CD4ϩCD28 null T cells are functional and are not anergic as might be expected (21). T cell receptor cross-linking either by immobilized anti-CD3 or soluble anti-CD3 and monocytes has been found to elicit proliferation. Under these stimulation conditions, these cells also produce high levels of cytokines, among which are IL-2, IL-4, and ␥-interferon. However, despite the ability of these unusual T cells to respond to T cell receptor-mediated stimuli, FIG. 8. Inactivation of the minimal CD28 promoter by mutations of the DNA-protein interaction sites ␣ and ␤. Clusters of nucleotide substitutions, as indicated, were introduced into the minimal CD28 promoter construct 42 shown in Fig. 3), cloned into a luciferase reporter vector, and used in transient transfection assays with Jurkat cells. Data shown are the mean (Ϯ S.E.) luciferase activities of three separate but concurrent transfections for each of two clones of the wild type (WT) or mutant (M1-M11) constructs. Luciferase activities are expressed as fold induction over the minimal SV40 promoter-driven control luciferase reporter pGLPromoter. The specific reporter activities were normalized against a cotransfected reporter, the CMV-driven ␤-galactosidase plasmid pCMV␤gal (32). For these experiments, a CMV-driven luciferase reporter, pCMVLuc, showed activities over 100-fold above the control, as expected (data not shown). Luciferase and ␤-galactosidase activities were determined by bioluminescence. Data shown are representative of two independent experiments involving triplicate concurrent transfections. CD4ϩCD28 null T cells fail to express CD40 ligand and are unable to support B cell proliferation and immunoglobulin production in vitro (22). This corroborates previous studies that demonstrated the requirement of CD28 in T cell-dependent B cell-mediated immune responses (43). Along these lines, we propose that accumulation of CD4ϩCD28 null T cells with age may contribute to the well-documented phenomenon of humoral hyporesponsiveness among the elderly. As epidemiologic data indicate, low responsiveness to vaccination is not uncommon among elderly individuals. For instance, the effectiveness of influenza vaccine is markedly reduced in the elderly as compared with younger individuals (13,14).
Delineation of cis-Acting Elements and Identification of Sequence Motifs Associated with CD28 Deficiency-Inasmuch as the regulatory pathways involved in the normal expression of CD28 are not yet known, the present work provides basic information on the minimal structural requirements for CD28 gene transcription. Here, we have cloned 600 bp of the 5Ј-UTR and provide evidence that the proximal 400 bp contain the minimal cis-acting elements necessary for gene expression as assessed by reporter gene bioassays. As the data show, only reporter plasmids containing the proximal 400 bp yielded promoter activity. Structural analysis of the 5Ј-UTR (Fig. 7) revealed that this proximal 400-bp segment contained the putative CAAT and TATA boxes and various constituent binding sites where the RNA polymerase complex presumably assembles (44,45). Deletion of this region, as shown by transfection experiments with reporters containing only the most proximal 200-bp segment (Fig. 3B), did not yield any significant promoter activity. Conversely, deletion of the proximal 200 bp also did not yield significant promoter activity, indicating that the TATA-containing region requires downstream elements to function.
Flanking the 400-bp region is a 200-bp segment that appeared to be inhibitory to promoter activity. The reason for this inhibition is not clear. However, results of the structural analysis (Fig. 7) revealed the presence of sequences within this distal 200-bp segment that are homologous to repressor elements found in other genes.
As shown previously, CD28 expression transiently increases upon T cell activation, and ligation of CD28 by specific antibody induces reduction in its surface expression (16,29). These findings suggest that there are regulatory elements responsive to T cell activation signals that are involved in feedback control. Our data indicate that such elements may be contained within the proximal 400-bp minimal promoter. Cells transfected with reporter constructs under the control of this minimal 400-bp promoter showed significantly down-regulated reporter activity when exposed to ionomycin and PMA (Fig. 3C).
An important finding of the present study was that the promoter activity of the proximal 400-bp segment of the 5Ј-UTR is suppressed in HUT78 cells (Fig. 3D). This CD4ϩ T cell lymphoma has been shown to lack CD28 at both the mRNA and protein levels without apparent genomic lesions in the CD28 gene itself (27). This observation, together with the fact that reverse transcription-PCR analysis on a variety of untransformed CD4ϩCD28 null T cell lines and T cell clones showed the uniform absence of CD28 transcripts, 2, 3 supports the notion that deficiency in CD28 expression may be due to the lack of relevant trans-acting factors.
This hypothesis was evaluated by EMSAs with nuclear extracts from CD28ϩ and CD28 null CD4ϩ T cells. The oligonucleotide probes used encompass a 67-bp segment containing putative binding sites for the Elk1, Ets1, and AP3 transcription factors (Figs. 4 and 7). As the data show, there are two noncompeting DNA binding activities within this sequence for extracts from CD28ϩ but not CD28 null CD4ϩ T cells. These binding activities correspond to two nucleotide stretches that lie on opposite flanks of the Ets1 motif (Figs. 4 -7). The 5Ј site, referred to as site ␣, is defined by probes 41A and 4441A and spans positions Ϫ206 to Ϫ179 of the minimal promoter. The 3Ј site, referred to as site ␤, is defined by probes 41B and 5152 and spans positions Ϫ171 to Ϫ148 of the minimal promoter.
Although site ␣ overlaps with an Elk1, the observed binding activity does not appear to involve Elk-and Ets-related transcription factors as determined by three lines of evidence. First, there is a lack of any detectable binding activity with a subsite of the 67-bp sequence containing the Ets motif (Figs. 4 and 5, probe 41C). Moreover, similar assays using binding probes containing Ets-and Elk-specific sequences previously characterized for other genes showed negligible binding activities. Second, the observed binding activities are not recognized by antibodies to Ets1, Elk1, and Elf, the three predominant lymphoid cell-specific members of the Ets family of transcription factors (36, 46 -50). Therefore, the direct binding of Etsrelated factors to DNA does not account for the observed binding activities, but the possibility that Ets-related factors are a component of such activities cannot be totally ruled out. Previous studies show that Ets family members may form complexes with DNA-binding proteins (37,51,52). Hence, it is possible that such interactions can bury antibody epitopes. Third, results of the mutational analysis show that nucleotide positions that abrogate factor binding activity of site ␣ are situated 3Ј of the Elk1-like sequence (Figs. 8 and 9), strong evidence that the latter may not be involved in DNA-protein complex formation.
Site ␤, on the other hand, overlaps with an AP3-like sequence. However, it is also unlikely that the observed differential binding activities of site ␤ for nuclear extract from CD28ϩ and CD28 null CD4ϩ T cells represent an AP3 or an AP3-like factor. AP3 has not been documented to play any role in T cell-specific gene transcription, although it has been shown to be important in the differentiation of myeloid cells (41). An AP3-like transcription factor has also been implicated in the inducible expression of a novel mast cell-derived cytokine (42). In the present study, known AP3-like sequences did not compete with the binding activities of the CD28 67-bp sequence. In a manner similar to that of site ␣, DNA-protein complex formation with site ␤ is also abrogated by mutations situated 3Ј of the overlapping AP3-like sequence (Figs. 8 and 9), indicating that AP3-like factors may not be involved.
Taken together, the data presented are consistent with the hypothesis that the deficiency in CD28 expression in CD4ϩ T cells is associated with the lack of relevant trans-acting factors, among which are those specific for sites ␣ and ␤. The observed binding activities of these two sites are clearly correlated with the CD28 phenotype and are uniformly found among CD4ϩ T cells. The loss of binding activities of sites ␣ and ␤ due to mutations strongly correlates with the inactivation of the minimal CD28 gene promoter (Figs. 8 and 9), indicating that these two sites are critical to the initiation of transcription. The requirement of sites ␣ and ␤ for promoter activity is further supported by the findings that neither the segment containing the CAAT and TATA boxes (Ϫ400 to Ϫ200 bp of the 5Ј-UTR) nor the proximal 200-bp region (containing sites ␣ and ␤) by themselves exhibit promoter activity (Figs. 3B and 7). These observations strongly indicate that the factors bound by sites ␣ and ␤ are required for the initiation of transcription.
The 67-bp segment containing sites ␣ and ␤ overlaps with the presumed transcription initiation site previously reported by Lee et al. (28). As noted by the authors, the reported length of the CD28 transcription initiation site is rather imprecise because of structural constraints brought about by upstream Alu sequences. Thus, the reported transcription initiation site is, at best, a partial map.
Regardless of the actual length of the transcription initiation site, sites ␣ and ␤ are contained within this 67-bp region and are situated immediately downstream from the putative TATA box (Fig. 7). The significance of this unusual topographic arrangement is not known at this time, but at least two hypotheses may be postulated as to the role the two sites in CD28 gene transcription. One hypothesis is that sites ␣ and ␤ may comprise the equivalent of an Initiator element (Inr) described in other genes (45,53,54). It is thought that such elements interact with TATA-bound complex and are critical in directing synthesis of initiated transcripts (45). Inrs may overlap with or encompass the entire transcription initiation site (45), as has been found in the promoters of eukaryotic initiation factor 2␣, terminal deoxynucleotide transferase, dihydrofolate reductase, and adenovirus major late protein (54). The common feature of Inrs is that mutations can block transcriptional activation (45,53,54). In this regard, we show in the present work that mutations within sites ␣ and ␤ do inactivate the minimal CD28 promoter (Fig. 8). A structural analysis of sites ␣ and ␤ (data not shown), however, reveals the lack of sequence homology with the known families of Inr (54).
An alternative hypothesis is that sites ␣ and ␤ bind factors that either are components of or are independent activators of the TATA-bound RNA polymerase complex. Although many of the core protein components of the TATA-bound complex have been described, it is possible that there are yet uncharacterized additional components (44,45). It is conceivable that among these components are cell-or tissue-specific factors. Along these lines, we currently have evidence that sites ␣ and ␤ binding activities are seen only with lymphoid tissues and with various T cell lines that express CD28 on their surface. 4 Whether the site ␣and ␤-bound factors are initiators or gene-specific components of the TATA-bound protein assemblage, our data consistently demonstrated their critical role in the the promoter activity of the CD28 5Ј-UTR. The identification of the factors bound by the two sites may be among the key information that will further elucidate the biochemical pathway for CD28 gene transcription. Ultimately, the identification of the genes encoding for ␣and ␤-binding proteins will facilitate our understanding of the origin, fate, and biology of CD4ϩCD28 null T cells found in elderly individuals.