Functional disruption of the CD28 gene transcriptional initiator in senescent T cells.

We recently reported that aging is accompanied by the emergence of CD4(+)CD28(null) T cells, a functionally aberrant lymphocyte subset rarely seen in individuals younger than 40 years. Here, we directly examined whether the lack of CD28 expression is due to a defect at the level of transcriptional initiation. Molecular studies reveal that CD28 gene transcription is controlled by two sequence motifs, sites alpha and beta. In vitro transcription assays using initiator-dependent DNA templates revealed that reversed polarity or the deletion of either motif inhibited transcription, indicating that alpha/beta sequences constitute a composite initiator. Moreover, nuclear extracts from CD28(null) cells failed to activate transcription of alphabeta-initiator DNA templates. Transcription of such templates was, however, restored with the addition of extracts from CD28(+) cells. Although previously described initiator elements have been defined by a consensus sequence, the alphabeta-initiator has no homology to such sequence. These studies demonstrate that initiators have functions other than positioning elements for the basal transcription complex. Rather, initiators can have a direct role in regulating the expression of specific genes. The gain or loss of initiator activity can be an important determinant of cell phenotypes.

The cellular and molecular processes underlying immune dysfunctions during normal aging are complex. The immune system undergoes a constant turnover of cells and is highly dependent on the replenishment of new precursor cells. However, de novo production of T cells rapidly declines with the progressive involution of the thymus with age (1,2). Consequently, there is replicative stress resulting in the progressive shortening of telomeres of peripheral lymphocytes (3,4). Replicative senescence is associated with altered patterns of gene expression (5). Among T cells, senescence is accompanied by a characteristic loss of CD28, predominantly among CD8 ϩ T cells (6,7) and to a lesser degree among CD4 ϩ T cells (8). Interestingly, CD4 ϩ CD28 null T cells have also been found in patients with chronic inflammatory syndromes, such as those seen in rheumatoid arthritis, Wegener's granulomatosis, and coronary artery disease (9 -11). A CD28 null phenotype is stable. Neither the triggering of the T cell receptor (TCR) 1 nor signals gener-ated by pharmacologic agents such as phorbol ester and calcium ionophore that bypass the TCR can restore CD28 expression (8,12,13).
Because CD28 is the dominant costimulatory molecule required for T cell activation, proliferation, and effector function (14), elucidation of the molecular basis for CD28 deficiency is of paramount interest. Inasmuch as CD28 null T cells uniformly lack specific mRNA of all known splice variants (8,12,15), we evaluated the hypothesis that the loss of CD28 is due to a transcriptional block. In previous work, we reported that CD28 expression is controlled by two sequences, sites ␣ and ␤, that are surprisingly situated immediately downstream from the TATA box (8). Nuclear proteins that specifically bind to sites ␣ and ␤ are limited to lymphoid tissue, and their expression patterns are correlated with the presence or absence of CD28 on the surfaces of T and B cells (15). Moreover, random mutations in either site can sufficiently inactivate promoter activity in reporter gene bioassays (8). The functional relevance of these sequence motifs is further indicated by the modulation of ␣/␤nuclear protein binding profiles by TCR triggering and during replicative senescence, two conditions that induce down-regulation of CD28 on the T cell surface (15).
Our finding that sites ␣ and ␤ map downstream from the TATA box (8) suggests that the expression of CD28 might be controlled at the level of transcriptional initiation. Although the TATA box has been traditionally considered as the assembly site of the basal transcription complex, flanking initiator (INR) sequences have been found to be critical core promoter elements (16 -19). In TATA-containing promoters, such INRs are thought to be positioning elements that tether TATA-binding protein, ensuring the fidelity of transcription from the TATA box (20). INR-binding proteins may also interact with TATA-binding protein-associated factors, resulting in improved efficiency of transcription, as has been demonstrated for immunoglobulin promoters (21,22). Although the existence of distinct INR-binding proteins is not clear, several proteins have been implicated in the activity of INRs. These include general transcription factors such as transcription factor II-I (23,24) and the components of transcription factor IID (25)(26)(27), or regulatory proteins such as YY1 (28) and USF (29,30). In the present work, the role of sites ␣ and ␤ as INRs was evaluated. Although these sequences have no homology with the consensus INR (18), sites ␣ and ␤ coincide with the putative transcription start site (31).

EXPERIMENTAL PROCEDURES
Cell Culture-T cell lines and clones were established from peripheral blood as described previously (8,15,32). Briefly, CD4 ϩ CD28 null and CD4 ϩ CD28 ϩ T cells were isolated from blood mononuclear cells by standard fluorescence-activated cell sorting procedures. Cells were * This work was supported by the Mayo Foundation and National Institutes of Health Grants RO1-AG15043 and RO3-AR45830. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  stimulated with anti-CD3 (OKT3, ATCC, Manassas, VA) and ␥-irradiated autologous monocytes for 24 h. Subsequently, cells were subjected to limited dilution cloning in 96-well plates with feeder cells consisting of ␥-irradiated, neuraminidase-treated EBV-transformed B lymphoblastoid cells without additional stimulation. Clones were isolated, and phenotypes were ascertained by immunofluorescence staining and flow cytometry (see below). Primary CD4 ϩ T cell lines were derived from unfractionated blood mononuclear cells that were similarly stimulated with anti-CD3. After 24 h, CD4 ϩ T cells were isolated by the immunodepletion of CD8 ϩ cells using the VarioMacs system (Biotec Miltenyi, Auburn, CA). Purity of the isolated cells was verified by flow cytometry. Cells were cultured on EBV-transformed B cell feeders and 20 units/ml recombinant human interleukin 2 (Proleukin, Chiron, Emeryville, CA). Sublines of CD4 ϩ CD28 ϩ and CD4 ϩ CD28 null T cells were subsequently established by fluorescence-activated cell sorting. All lines and clones were maintained by weekly stimulation with EBV-transformed B cell feeders and recombinant human interleukin 2 in a humidified 7.5% CO 2 incubator as described previously (8,15).
The T cell lymphoma lines Jurkat and HUT78 (ATCC) were maintained at densities of about 5 ϫ 10 6 cells/ml in RPMI 1640 medium supplemented with 10% fetal calf serum. HUT78 was cultured in the presence of 20 units/ml recombinant human interleukin 2. Cells were maintained in a humidified 5% CO 2 incubator.
Phenotyping of Cells-T cells were examined for cell surface expression of CD3, CD4, and CD28 by three-color immunofluorescence staining and flow cytometry. The lack of CD28 expression in T cell clones or lines was also verified by reverse transcription-polymerase chain reaction (reverse transcription-PCR) procedures for the splice variants of CD28 (12,15).
Clonality of the T cell clones was established by standard nested reverse transcription-PCR for the BV-BJ segments of the TCR. PCR products were cloned into the TA vector and recombinants were used to transform One-Shot TM Escherichia coli (Invitrogen, Carlsbad, CA). Sequencing of plasmids prepared from at least three randomly selected bacterial colonies authenticated the clones.
Nuclear Extracts-Nuclear extracts were prepared using a high salt extraction protocol described previously (8,33). Extracts from the primary T cell lines and clones were prepared between 3 and 5 days after the last stimulation. Extracts from Jurkat and HUT78 cells were prepared during logarithmic growth. Protein concentrations of the extracts were determined by the Bradford method using a protein assay kit (Bio-Rad). Nuclear extracts were aliquoted, snap frozen in liquid nitrogen, and stored at Ϫ70°C.
In Vitro Transcription Assay-Sequences of the human CD28 gene sites ␣ and ␤, CGTTATATCCTGTGTGAAATGCTGCAGTCAGGATGC-CTTGTGGTTTGAGTGCCTTGAT (the underlined 5Ј and 3Ј sequences correspond to ␣ and ␤, respectively) (8) were cloned into plasmid templates (provided by Dr. Jörg Kaufmann, Chiron Corp.) containing a 180-base pair G-less cassette downstream from a consensus TATA and the INR of terminal deoxynucleotidyl transferase (TdT) (34,35). Sites ␣ and ␤ were introduced into these plasmids as separate elements or as a contiguous unit replacing TdT-INR by the gene soeing technique (36). Templates containing a reversed orientation of ␣␤ were also made. Constructs were amplified in E. coli DH5␣ (Life Technologies, Inc.) by standard transformation procedures and randomly selected bacterial colonies were screened for recombinant plasmids by PCR using primers specifically designed to detect the inserted ␣ and/or ␤ sequence. Where PCR amplification of ␣␤ was indicated, plasmids were prepared by a commercially available kit (EndoFree plasmid kit, Qiagen, Valencia, CA) and subjected to DNA sequencing of the entire the region spanning TATA, INR, and the G-less cassette. Two clones of each construct were selected for these studies.
Conditions of transcription reactions were as described previously (34) with the following modification. Nuclear extracts were subjected to centrifugation dialysis (Microcon YM3 Amicon filter, Millipore, Bedford, MA) against 10 volumes of reaction buffer, and the total protein concentration was determined as above. Nuclear extracts were added at the indicated amounts to 300 ng of plasmid template and incubated at 30°C for 60 min. A mixture of 100 mM ATP, 100 mM CTP, and 50 M [␣-32 P]UTP (Amersham Pharmacia Biotech) was added, the total reaction volume was adjusted to 100 l, and the mixture was incubated for 90 min at 30°C. Transcription products were digested with 60 units of RNase T1 (Roche Molecular Biochemicals), extracted with phenol-chloroform, size-fractionated on 8% polyacrylamide-6 M urea sequencing gels, and visualized by autoradiography.
Verification of ␣␤-Specific and TdT-INR Activating Proteins-Nuclear extracts used in the in vitro transcription assays were tested for the presence of ␣and ␤-binding factors by gel shift assays by previously described procedures (8,15). Centrifugation dialysis of extracts (see above) did not significantly affect the ␣/␤ binding profiles of the extracts (data not shown). As in previous studies, CD28 ϩ T cells consistently showed ␣and ␤-specific binding proteins. Additionally, competitive gel shift assays with ␣, ␤, or TdT-INR sequences revealed exquisite specificity of binding activities among these 3 promoter motifs, as they did not cross-compete each other (data not shown). Other studies indicate similar distinctiveness of the binding activities of TdT-INR from those of other INR sequences (37).
Although the transcription factor that directly bind TdT-INR is not known, transcriptional activation of the TATA/TdT-INR template has been shown to be dependent on CIF150, a cofactor of transcription factor IID that appears to be ubiquitously expressed (35). Reverse transcription-PCR experiments for CIF150 expression in T cells used in the present study uniformly showed the presence of specific transcripts. Direct sequencing of PCR products (data not shown) authenticated these CIF150 transcripts.

RESULTS
Sites ␣ and ␤ Constitute a Bona Fide INR Element-We have shown previously that two sequence motifs, sites ␣ and ␤, regulate the constitutive expression of CD28 (8). The peculiar topographic location of these sequences immediately flanking the TATA box suggested that they might directly interact with the basal transcription complex. To evaluate this hypothesis, we adapted an in vitro transcription system (35) that measures transcription of a DNA cassette controlled by a consensus TATA box and TdT-INR. As shown in Fig. 1, replacement of the TdT-INR element of the DNA template by the CD28 gene ␣␤ sequence resulted in the production of cassette transcripts in the presence of nuclear extracts from Jurkat, a T cell lymphoma that expressed high levels of CD28 (15). The levels of ␣␤-dependent transcription increased with the amounts of nuclear extract added in a manner similar to those seen with templates containing the TdT-INR. Consistent with previous studies (34,35), the mutated variant of the TdT-INR yielded levels of cassette transcripts that were consistently and significantly lower than that seen with wild-type TdT-INR. Presumably, the low amounts of transcripts seen with the mutant TdT-INR represented the basal level of transcription from the upstream canonical TATA box.
Experiments were also carried out to examine whether sites ␣ and ␤ can function independent of each other. As shown in Fig. 2, neither ␣ nor ␤ alone elicited transcription at levels higher than the baseline seen with templates containing the TdT-INR mutant. As in the previous experiment, the composite ␣␤ sequence elicited high levels of transcription in the presence of Jurkat nuclear extracts in a dose-dependent manner. The amounts of ␣␤-driven transcripts were equivalent to those seen with the wild-type TdT-INR.
In TATA-containing promoters, INRs are known to synergize with the TATA box, resulting in levels of transcription that are significantly higher than those seen with INR or TATA alone (38). This synergy is distinguished from a classical enhancer in that the latter induces transcription regardless of its orientation or distance from the basal complex assembled on the TATA box. Thus, we examined whether a reversed polarity, from ␣ 3 ␤ to ␤ 3 ␣, affects INR activity. As shown in Fig. 3, two independent clones of DNA templates containing the reversed ␤ 3 ␣ sequence indeed yielded only low amounts of cassette transcripts. In these template constructs, the levels of transcription were equivalent to those produced by constructs containing the mutant TdT-INR.
␣␤-INR in CD28 null T Cells Is Nonfunctional-In previous studies, ␣/␤-specific complexes were found to be uniformly lacking in CD4 ϩ CD28 null T cells. Gel shift assays revealed that neither contiguous ␣␤ (8) nor separate ␣ and ␤ (15) probes showed protein binding activities with nuclear extracts from CD28 null cells. Therefore, we examined whether this lack of DNA-protein complexes correlates with the absence of transcriptional activity. As shown in Fig. 4, in vitro transcription assays using nuclear extracts from various activated CD28 ϩ T cells yielded cassette transcripts from DNA templates containing ␣␤ as the INR element. The amounts of transcripts produced were equivalent to those seen with Jurkat nuclear extracts. In contrast, none of the nuclear extracts from CD28 null T cells elicited transcription above basal levels. Extracts from CD28 ϩ and CD28 null T cells were, however, indistinguishable in their ability to promote transcription of templates containing the TdT-INR. As expected, templates with the mutated TdT-INR produced equivalent low/basal amounts of transcripts regardless of the CD28 phenotype of the extracts used.
The lack of ␣␤-INR activity in CD28 null T cells could be due to the absence of ␣␤-specific transcription factors or to the presence of an inhibitor of ␣␤-proteins. To address this issue, reconstitution experiments were conducted using nuclear extracts from Jurkat and HUT78 cells as prototypes of CD28 ϩ and CD28 null T cells, respectively. As shown in Fig. 5, transcription of two DNA templates containing the ␣␤-INR was at low/basal levels with HUT78 extracts. However, the addition of increasing amounts of Jurkat extracts effectively restored ␣␤mediated transcription. The reconstitution of transcriptional activity was proportional to the amounts of Jurkat extracts added to the reaction. Such mixtures of Jurkat and HUT78 extracts did not alter the levels of transcription of templates containing TdT-INR.
Reciprocal experiments were also conducted wherein HUT78 extracts were added in increasing amounts to a constant amount of Jurkat extracts. As shown in Fig. 6, the addition of HUT78 extracts did not affect ␣␤-driven transcription in Jurkat extracts. Transcription from TdT-INR templates were also unaffected by the titration of HUT78 extracts. specific Transcription Factors-The observation that ␣␤-INR activity required ␣ and ␤ sequences in tandem (Fig. 2) suggested that nuclear proteins binding to both motifs are essential to transcription. Although these ␣/␤-binding proteins remain to be identified, previous studies indicate that ␣and ␤-complexes are distinct from each other (8). Therefore, we examined the effect of depletion of either complex in the efficiency of transcription. As shown in Fig. 7, incubation of Jurkat nuclear extracts in oligonucleotides corresponding to either site ␣ or ␤ resulted in the significant reduction of transcription of DNA templates containing the composite ␣␤ sequences as INR. The levels of ␣␤-driven transcription were effectively reduced to basal levels at oligonucleotide concentrations of 300 fmol. Similarly, incubation of extracts in oligonucleotides containing both ␣ and ␤ motifs also abrogated ␣␤-mediated transcription. As expected, none of the ␣/␤ oligonucleotides affected transcription of templates containing the TdT-INR. DISCUSSION The present work provides functional evidence for the direct role of sites ␣ and ␤ in CD28 gene transcription. Because the assay system adapted in this study stringently gauges the induction of INR-driven transcription over the basal activity of a canonical TATA box (34,35), the ability of ␣ and ␤ to function as a core promoter element in a heterologous DNA template ( Fig. 1) is impressive. Moreover, the lack of INR activity in similar templates with either sequence motif isolated from the other, as well as in templates with reversed polarity from an ␣ 3 ␤ to a ␤ 3 ␣ orientation (Figs. 2 and 3), provides compelling evidence that ␣ and ␤ constitute a singular INR. Incidentally, the sequence stretch encompassing ␣/␤ overlaps the previously described, albeit equivocated, transcriptional start site of the CD28 gene (31). The present data therefore authenticate this initiation site coinciding with ␣␤. Its peculiar location immediately flanking and downstream of the TATA box (8) lends further structural definition of ␣␤ as an INR in a manner similar to other known INRs (25,39,40).

␣␤-INR Activity Requires Coordinate Expression of Motif-
Although INRs serve as positional elements in TATA-containing promoters (20), they are the nucleation sites for the basal transcriptional complex in TATA-less promoters (41)(42)(43)(44). This is exemplified by the expression of lymphoid-specific genes such as TdT, the CD4 antigen, and the TCR V␤ chain, all of which lack a TATA box and utilize INR to initiate transcription (45)(46)(47). Interestingly, mutations in INRs have been associated with certain diseases. For example, the lack of ␤-globin expression in some forms of ␤-thalassemia has been associated with mutations in the INR of the gene (48). The use of such naturally occurring as well as synthetic mutants of ␤-globin INR in transcription assays in vitro do in fact reveal down-regulation or complete inhibition of transcription (49), providing physiological evidence for a direct link between INR function and cell/tissue phenotype.
There are two curious features of the ␣␤-INR of CD28. The first is that ␣␤-INR is a much longer element (8) compared with the loosely defined consensus INR sequence Py-Py-A ϩ1 -N-T/A-Py-Py (18) with which it has no homology. Although there are indications that this approximate consensus might be conserved among eukaryotes (50,51), there is an increasing body of evidence for INRs with divergent sequences. Among the evidence are the INRs of retinoic acid receptor ␤2 (37), mRNA cap-binding protein eIF4E (52), HIV-1 long terminal repeat (53), somatostatin receptor II (54), and vascular endothelial growth factor receptor (55). The second feature of ␣␤-INR is that sites ␣ and ␤ are nonoverlapping binding sites of discrete protein complexes (8,15). Although the binding of motif-specific transcription factors occurs independently, the cooperative interaction of ␣and ␤-bound proteins is required for transcriptional initiation. Indeed, the depletion of either ␣or ␤-binding proteins from nuclear extracts by decoy motif-specific oligonucleotides abolish ␣␤-INR activity (Fig. 7).
These peculiar properties of ␣␤-INR suggest that it may be a novel core promoter element. Interestingly, previous studies showed that ␣and ␤-binding factors are found only in lymphoid tissue (15) supporting the notion that ␣␤-INR might account for the restricted expression of CD28 to T cells and some transformed B cells. Although the INRs are primarily involved in enhancing nucleation of the basal transcription complex on the TATA box (20,56), there is also evidence for their accessory role in regulating expression of specific genes. A classic example is the regulation of the Drosophila alcohol dehydrogenase gene (57,58) in which the INR can discriminate between two tandem promoters that are used differentially at various stages of development. Cell-specific gene expression is also increasingly indicated to be INR-dependent. The lymphocyte specificity of TdT is INR-dependent (46). The presence of cell typespecific INR-binding proteins, albeit unidentified, have been implicated in the maximal activation of the human chorionic somatomammotropin promoter (59), the interferon-responsive promoters such those of Fc␥ receptor 1b (60), guanylate-binding protein, and H-2L d (61), and the cell-specific induction of the ␤2 isoform of the retinoic acid receptor (37). Because of the divergent sequences of INRs of these latter genes and the ␣␤-INR from the consensus sequence, it is quite possible that cell type-or gene-specific INR-binding proteins profoundly influence the programs of gene expression. Thus, the identification of ␣␤-INR transcription factors is pivotal to understanding The present data also unequivocally demonstrate that a CD28 null phenotype is related to the complete disruption of transcription. Because our assay system directly assessed the ability of ␣␤ to initiate transcription of a heterologous DNA template, the finding that the nuclear extracts from CD4 ϩ CD28 null T cells did not promote transcription of ␣␤driven templates is a compelling evidence for functional disruption of INR activity in these cells (Fig. 4). Such transcriptional incompetence of the CD28 null extracts could be restored by the addition of extracts from CD28 ϩ cells (Fig. 5), indicating the absence of transcription factor complexes that specifically recognize ␣␤-INR. This interpretation is supported by two other observations. The first is that excess amounts of CD28 null extracts do not perturb the transcriptional competency of extracts from CD28 ϩ cells (Fig. 6), hence the exclusion for the role of ␣␤-specific repressors. The second is that efficiency of transcription of TdT-INR driven templates are equivalent for both extracts (Figs. 4 -6). These findings corroborate previous data demonstrating the uniform and coordinate lack of ␣and ␤-bound complexes in CD4 ϩ CD28 null T cells (8,15). Whether or not the emergence of these unusual cells during normal aging (6 -8) or in chronic inflammatory diseases (9 -11) could be solely attributed to the down-regulation (or complete lack) of expression of a unique ␣␤-INR-binding protein(s) is not known at this time. Previous data, however, demonstrate that ␣/␤binding proteins are seen only in lymphoid tissues and that their binding activities are modulated in conditions known to induce changes in the levels of cell surface expression of CD28 (15).
It is important to note that CD4 ϩ CD28 null T cells, whether isolated from elderly individuals (8) or from patients with chronic inflammatory conditions (9,11), are functionally active (62,63) and not anergic as might be predicted from studies in the mouse (14). However, they have a perturbation in the pattern of expression of an array of molecules known to be important for lymphocyte function. For instance, CD4 ϩ CD28 null T cells also lack expression of CD40 ligand, hence are unable to support B cell proliferation and differentiation (13). They are highly resistant to apoptosis (32,64), which may explain their persistence in the circulation for years and their expansion up to 50% of the total CD4 compartment (8,9). A key question then is whether reconstitution of CD28 ␣␤-INR function can restore CD28 expression and consequently reestablish normal T cell effector function. Thus, the biochemical dissection of ␣␤-INR activity is of significant interest.
In conclusion, data presented here show that CD28 deficiency in CD4 ϩ T cells is due to a transcriptional block. Specifically, the ␣␤-INR element is nonfunctional because of a coordinate lack of ␣␤-specific transcription factors. This is unlike the well documented active repression of INR activity such as the INR-dependent, c-myc-mediated down-regulation of many genes (65)(66)(67). These studies collectively support the notion that INRs can have specific regulatory role in gene expression in addition to the nucleation of the basal transcription complex (20,56). In a manner similar to enhancers, such a regulatory role of INRs profoundly influences cell phenotype and function.