Molecular Basis for the Loss of CD28 Expression in Senescent T Cells*

CD28 null T cells are the most consistent biological in-dicator of the aging immune system in humans and are predictors of immunoincompetence in the elderly. The loss of CD28 is the result of an inoperative transcriptional initiator (INR), which consists of two nonoverlap-ping (cid:1) and (cid:2) motifs that have distinct protein binding profiles but function as a unit. In CD28 null T cells, there is a coordinate loss of (cid:1) -/ (cid:2) -bound complexes, hence the (cid:1)(cid:2) -INR is inactive. In the present work therefore, studies were conducted to identify the components of such complexes that may account for the trans -activation of the (cid:1)(cid:2) -INR. By affinity chromatography and tandem mass spectrometry, two proteins, namely, nucleolin and the A isoform of heterogeneous nuclear ribonucleopro-tein-D0 (hnRNP-D0A), were identified to be among the key components of the site (cid:1) complex. In DNA binding assays, specific antibodies indicated their antigenic presence in (cid:1) -bound complexes. Transcription assays showed that they are both required in the trans- activation of (cid:1)(cid:2) -INR-driven DNA templates. Because CD28 is

The CD28 molecule is a T cell-restricted membrane glycoprotein that provides the requisite costimulatory signal for the induction and maintenance of T cell-mediated immune responses (1,2). Coengagement of CD28 with the T cell receptor enhances the synthesis of several humoral growth factors including interleukin-2 and of anti-apoptotic molecules (3,4). Hence, T cells either become anergic or undergo apoptosis in the absence of CD28 signals. Targeted deletion of the CD28 gene in laboratory mice has been found to result in immunocompromised animals because of defective T cell activation (5)(6)(7). These findings underscore the central role of CD28 in adaptive immunity.
Although CD28 is constitutively expressed on all T cells, CD28 null T cells are typically found in the immune system of the elderly, in both CD8 ϩ (8,9) and CD4 ϩ compartments (10). CD28 null cells have highly shortened telomeres compared with their CD28 ϩ counterparts, indicating their long replicative history (11). These unusual cells are also highly oligoclonal (9,12), occurring at large clonal sizes that contribute to the contraction of the T cell repertoire diversity. Because of the limited replicative lifespan of T cells (13), CD28 null cells are thought to be biological indicators of immunosenescence. Interestingly, CD28 null CD4 ϩ T cells have also been found in high frequencies among patients with chronic inflammatory conditions such as rheumatoid arthritis (14), Wegener's granulomatosis (15), and unstable coronary artery disease (16). In these pathological states, large clonal populations of these cells have been postulated to represent a pool of prematurely senescent T cells resulting from chronic immune activation (10,17,18).
The CD28 null T cell phenotype is generally stable and lacks specific transcripts of all the known splice variants of CD28 (18 -21) resulting from a transcriptional block. Our studies show that the basal transcription of the CD28 gene is regulated by two sequence motifs sites ␣ and ␤, in the gene promoter, situated downstream from an atypical TATA box (10). These sequences constitute a functionally singular transcriptional initiator (INR) 1 element (18). In reporter gene bioassays and in vitro transcription studies, mutation in or deletion of either motif is sufficient to inactivate the CD28 gene promoter. In CD28 null T cells, the ␣␤-INR is functionally inoperative because of the coordinate lack of sites ␣and ␤-specific transcription factors (18,20). Although INRs are classically defined as nucleation sites of the basal transcription initiation complex (22,23), these findings indicate that loss (or gain) of INR activity may also be a critical determinant of cell phenotype and function.
Because the CD28 ␣␤-INR has no homology with other INRs (18,24,25), we undertook studies to characterize the relevant INR-binding proteins. We utilized a combination of affinity chromatography in concert with matrix-assisted laser desorption ionization-time of flight-mass spectrometry (MALDI-TOF-MS) and nanocapillary liquid chromatography-nanospray tandem mass spectrometry (nLC-MS/MS) to identify the proteins associated with the DNA-binding complex. By these approaches, we identified two of the component proteins of the transcription factor complex which recognize site ␣ of the CD28 INR. Here, evidence is also presented that the specific removal of either protein component of the site ␣-binding complex effectively inhibits the trans-activation of ␣␤-INR-driven DNA templates. The present work therefore provides a biochemical basis supporting the notion that the CD28 INR is indeed a structurally bipartite but functionally singular core promoter element.

EXPERIMENTAL PROCEDURES
Cell Culture and Nuclear Extracts-Jurkat (ATCC), a prototypical CD28 ϩ T cell line (18), was propagated in RPMI 1640 medium supplemented with 10% fetal calf serum at 37°C in a humidified 5% CO 2 incubator. Cells were maintained in batches of 250-ml flask cultures or 1-liter minibioreactors. Nuclear extracts were routinely prepared from bulk cultures when the cell density reached 5 ϫ 10 6 cells/ml. Nuclear extracts were prepared as described previously (10,20) and stored at Ϫ70°C until use. In these studies, a 75-liter culture was processed for nuclear protein extraction. Total protein concentration was determined colorimetrically using the Bio-Rad protein assay reagent.
T cells used in the present studies were routinely subjected to phenotypic screening by direct immunofluorescence staining for CD3, CD4, and CD28 and analyzed by flow cytometry as described previously (10, 18 -21). Expression, or lack thereof, of CD28 was also confirmed by reverse transcription (RT)-PCR assays using amplification primers for all of the known variants of CD28 as described elsewhere (20).
Affinity Chromatography-Nuclear extracts were dialyzed against 10 volumes of a hypotonic Hepes buffer pH 7.0 (10, 20) containing a protease inhibitor mixture (Roche Molecular Biochemicals), concentrated by centrifugation dialysis in Centricon YM-3 filters (Millipore), and subjected to sequential adsorption by column chromatography. The adsorption columns were agarose matrices of immobilized commercial DNA (Amersham Biosciences), heparin, and strepavidin (Pierce). The flow-through from the strepavidin column was subsequently poured into an affinity column with immobilized double-stranded, biotinylated, synthetic oligonucleotide corresponding to site ␣ of the CD28-INR or its mutated variant M3 (10). The affinity column was washed aseptically with 10 volumes of Hepes buffer in the absence of peptide protease inhibitors. Bound proteins were eluted with 2 M KCl and concentrated by centrifugation dialysis against 10 volumes of sterile 10 mM Hepes pH 7.0.
The simple design of the affinity purification for site ␣-binding proteins was based on empirical studies using microcolumns (data not shown). Additional adsorption/clearing steps such as the use of DEAE or Mono Q columns neither improved the elution yields nor altered the binding specificity of the site ␣ oligonucleotide affinity column. As expected, incubation of precleared nuclear extracts with excess amounts of soluble site ␣ oligonucleotides resulted in a significant reduction in the amounts, or the complete lack, of proteins that can be eluted from the affinity column. In these latter experiments, the absence of proteins eluted from the columns was confirmed by SDS-PAGE and silver staining and by MALDI-TOF-MS (below).
MS Studies-Eluates from DNA affinity columns were initially subjected to MALDI-TOF-MS to examine the relative diversity of the isolated site ␣-specific protein complexes. Samples were desalted on a C4 ZipTip cartridge (Millipore). Retained proteins were eluted with a 2-l matrix solution of 10 g/l 3,5-dimethoxy-4-hydroxycinnamic acid in 55% acetonitrile (ACN), 0.1% trifluoroacetic acid. Two 0.8-l aliquots were loaded onto a 0.5-l sinnapinnic acid matrix precrystallized in 70% ACN and 0.1% trifluoroacetic acid. Mass spectra were acquired using a Voyager-DE STR mass spectrometer (Perseptive Biosystems, Framingham, MA) by delayed extraction using either the reflectron or linear mode. Acceleration grid and guide wire voltages were set to 20,000 V, at 70% or 0.08%, respectively. The low mass gate was set to either 600 or 5,000. External calibration in linear mode was performed using doubly and singly charged ions from bovine serum albumin prepared by the same procedure as the sample eluates.
MS/MS was performed to identify the site ␣-specific proteins by examining the peptide fragmentation fingerprints of the affinity column eluates. Desalted samples (see above) were dissolved in 10 mM Hepes pH 8.0 to a maximum concentration of 2.5 g/l and subjected to cysteine reduction and alkylation. Dithiothreitol (in 1 M Tris-HCl pH 8.8) was added to the samples to a final concentration of 1 g/l and incubated for 30 min at 37°C. Samples were cooled to room temperature, iodoacetamide was added to a concentration of 2 g/l, and the samples were incubated for 30 min in the dark. Subsequent to reduction and alkylation, samples were diluted with an equal volume of 100 mM Tris-HCl buffer to a final concentration of 1.25 g/l and digested overnight with trypsin (E/S 1:50) at 37°C. Trypsin digestion was stopped with the addition of 10% formic acid. Aliquots of the trypsindigested material were diluted with an equal volume of 0.1% trifluoroacetic acid and loaded onto a ZipTip cartridge packed with C18 reverse material (Millipore) as described by the manufacturer. Peptides were eluted with 3 l of matrix solution (12 g/l ␣-cyano-4-hydroxycinnamic acid in 45% aqueous ACN and 0.1% trifluoroacetic acid).
Peptides were subjected to nLC-MS and MS/MS as described previously (26). Briefly, reversed phase scale LC separations were done on a prepacked 75-m inner diameter/5-cm long PicroFrit column (New Objective Inc., Cambridge, MA) packed with 5-m particles of Aquasil C18 (ThermoHypersil-Keystone, Bellefonte, PA). Peptides were eluted at a flow rate of 0.2 l/min utilizing a linear gradient as follows: initial hold at 0% B for 10 min, 1-min ramp to 10% B, 10 -50% B over 30 min, 50 -95% B over 5 min, hold 5 min at 95% B, return to 0% B over 5 min, and reequilibration for 5 min prior to new injection. Mobile phase A consisted of water/ACN/n-propyl alcohol (98/1/1 v/v/v) containing 0.2% formic acid. Mobile phase B consisted of ACN/n-propyl alcohol/water (80/10/10 v/v/v) containing 0.2% formic acid. Mobile phase flows at 50 l/min were supplied by a Michrom UMA LC system (Michrom Bioresources Inc., Auburn, CA). A contact closure event table within the LC software was used to send start signals to the autosampler, control the LC switching valve, and send acquisition start signals to the mass spectrometer. A 10-min (10 min ϫ 10 l/min ϭ 100 l) sample transfer and wash step was built into the beginning of the reversed phase scale LC method to allow reconcentration on the mPC disk (SDB-EX styrene/ divinylbenzene disk, Varian Inc., Harbor City, CA) while simultaneously reequilibrating the scale LC column from the previous gradient. At the conclusion of 10 min, the mPC disk was switched on-line with the LC column, the gradient started, and mass spectrometer data acquisition commenced.
Typically, 5-8 l of the digested protein sample was concentrated on the mPC membrane. Because the SDB material in the mPC disk is less rententive than the C18 LC column, it allowed analytes eluting from the membrane to be refocused briefly on the head of the LC column during the reversed phase gradient.
Nanospray ESI-MS was performed using a Micromass Q-TOF II (Micromass, Beverly, MA) equipped with a modified Micromass ESI source. The source was modified by replacing the mounting platform on the X/Y/Z manipulator with a 2-piece platform of stainless steel on top of insulating Delrin that contains a grid of mounting holes. A titanium microvolume union with 150-m bore (Valco, Houston, TX) was mounted to the platform via the bulkhead threads in the union and a nylon screw. The electrospray voltage, typically 1.7-2.1 kV, was applied to the metal microvolume union through the stainless steel mounting plate. The Delrin base of the platform serves to insulate electrically the spray platform from the rest of the X/Y/Z manipulator assembly. Spectra were acquired on either the MS or the auto MS/MS mode. Auto MS/MS experiments were conducted using survey scans to choose up to three precursor ions. Collision energies were chosen automatically as a function of m/z value and charge. Argon was used as the collision gas. The mass axis of the TOF analyzer was calibrated by manually injecting 0.3 l of 0.1 mg/ml NaI dissolved in isopropyl alcohol/water (50/50 v/v) through the LC column. The solution also contained a small amount of cesium ion allowing calibration over the m/z range 132.9054 -1821.7206 using a linear fit of the calibration points.
Data base searches were carried out using either accurate peptide masses or partial sequence information. Searches were performed utilizing the Protein Prospector search algorithm (prospector.ucsf.edu) MS-Fit or the Mascot search program (Matrix Science Limited, www.matrixscience.com), and the NCBI protein data base.
Electrophoretic Mobility Shift Assays (EMSAs)-Purity of the samples at each phase of column chromatography was monitored by EM-SAs, which were performed as described previously (10,20). As indicated, EMSAs were carried out with the addition of specific antibodies or the appropriate isotype control immunoglobulin (Ig) or preimmune antiserum to the binding reactions. In these studies, the anti-nucleolin monoclonal antibody MS3 (27) (provided by Dr. Ben Valdez, Baylor College of Medicine) and four rabbit antisera to heterogeneous nuclear ribonucleoprotein (hnRNP)-D0 (provided by Dr. Mate Tolnay, Walter Reed Research Institute) were used at the indicated dilutions. The specificities of these rabbit antisera to the A (P3, P4) and B isoforms (P1) or to a conserved region (P2) of hnRNP-D0 have been described previously (28,29).
Competitive EMSA was also carried out as described previously (10). Sequences corresponding to the Ig switch region recognized by a B cell-specific transcription factor LR1 (30,31), the hnRNP-D0B binding motif in the complement receptor 2 (CR2) promoter (28,32), or the mutated variant (M3) of site ␣ (10) were used as competitors to the CD28 site ␣ binding probe.
DNA binding assays were also conducted using LR1 and CR2 doublestranded oligonucleotide sequences as binding probes. In other assays as indicated, single-stranded oligonucleotides of site ␣ were also used as binding probes.
In Vitro Transcription Assay-Transcription assays with INR-driven DNA templates were conducted as described previously (18). CD28-INR-driven DNA templates contained either the wild type or mutated variants of site ␣ (M3, M4) or site ␤ (M9, and M10) (10). Mutants were generated by the gene splicing by overlap extension technique described elsewhere (33).
For assays using immunodepleted extracts, batches of 100-g aliquots of dialyzed nuclear extracts were incubated overnight at 4°C with saturating amounts of anti-nucleolin (MS3) or anti-hnRNP-D0A (P3, P4), or equivalent amounts of IgG or preimmune serum. To this mixture, protein A/G-agarose (Pierce) was added and incubated for another 6 h at 4°C. Supernatants were collected after brief centrifugation and concentrated by centrifugation dialysis in Microcon YM-3 filters (Millipore). Protein concentration was determined using the Bio-Rad protein assay reagent. Thirty g of each of the antibody-cleared extracts was used in transcription assays using CD28 ␣␤-INR-driven DNA templates.
In similar experiments, the immunoprecipitated protein complexes were added back the antibody-cleared extracts. Protein A/G-bound proteins from the depletion experiments were subjected to high salt elution, concentrated by centrifugation dialysis, and 20 g of the concentrate was added to the antibody-cleared extract and used in transcription assays.
DNA templates containing either the wild type or mutated form of the INR of the terminal deoxynucleotidyltransferase (TdT) gene (34) were also used as system controls.
Western Blotting-Nuclear extracts from CD28 ϩ and CD28 null T cells were prepared as described above, and 10-g aliquots were subjected to SDS-PAGE under reducing conditions in 10% polyacrylamide gels. Fractionated proteins were transferred to nitrocellulose membranes (0.2-m sieve, Bio-Rad) by standard electroblotting procedures. Membranes were blocked with 4% bovine serum albumin in Tris-buffered saline pH 7.4 for 1 h, followed by a 1-h incubation in a 1/1,000 dilution (in Tris-buffered saline containing 1% bovine serum albumin and 0.25% Tween 20) of the anti-nucleolin antibody MS3 (27) or the P4 rabbit antiserum to hnRNP-D0A (29). Membranes were washed extensively in the Tris-buffered saline-bovine serum albumin-Tween dilution buffer and subsequently incubated for 1 h in a 1:1/000 dilution of horseradish peroxidase-conjugated goat anti-mouse Ig (BD Biosciences) or goat antirabbit IgG/IgL (BIOSOURCE Intl., Camarillo, CA) for MS3-or P4treated membrane, respectively. The membranes were again washed extensively in Tris-buffered saline-bovine serum albumin-Tween dilution buffer, and the immunoblots were developed by chemoluminescence using the SuperSignal kit (Pierce).
Isolated site ␣-bound proteins were also subjected to Western blotting to ascertain the presence of nucleolin and hnRNP-D0A in the DNA⅐protein complexes. Approximately 100-g samples of nuclear extracts from Jurkat and HUT78 cells were incubated separately with 100 l of freshly washed strepavidin-agarose slurry (Pierce) for 2 h at 4°C. After a brief centrifugation, the precleared extracts were divided into two aliquots; one was left on ice until use, and the other was added to a 500-l EMSA binding reaction (as described above) containing 10 nmol of biotinylated site ␣ sequences and incubated on ice for 1 h.
A fresh 100-l slurry of strepavidin-agarose was washed three times with the reaction buffer. After the last centrifugation, the supernatant was discarded by vacuum aspiration, and the EMSA binding reaction was poured into the strepavidin-agarose pellet. The mixture was incubated for 1 h at 4°C in a rotating wheel. After a brief centrifugation, the supernatant was carefully aspirated off into a microfuge tube and left on ice. The DNA⅐protein complexes/strepavidin-agarose pellet was washed twice by centrifugation in 3 volumes of the binding reaction buffer. The binding reaction supernatant and the DNA-bound fraction, along with the precleared nuclear extract, were each mixed with an equal volume of 2ϫ Laemmli buffer, and 100-l aliquots were subjected to SDS-PAGE and Western blotting for nucleolin and hnRNP-D0A as described above. As system control, similar immunoblotting experiments were conducted using the monoclonal antibody 4B10 (35) (provided by Dr. Gideon Dreyfuss, HHMI, University of Pennsylvania), which specifically recognizes the RNA-binding protein hnRNP-A1 (36) but not the isoforms of hnRNP-D0.
RT-PCR Assay for hnRNP-D0A-Total RNA from a panel of T cells, as indicated, was prepared using the Trizol reagent (Invitrogen) and subjected to first strand cDNA synthesis by standard procedures. Aliquots of cDNA samples were subjected to PCR using specific primers. The sequences of the primer pairs used were gaggtggtggccccagt and cactctgctggttgctataatc, which amplified a 168-bp product corresponding to exon 7 of hnRNP-D0A (GenBank accession no. D55674; Ref. 37). PCR was carried out in 30 cycles of 94°C for 1 min, 55°C for 2 min, and 72°C for 2 min. PCR products were size fractionated by agarose gel electrophoresis and visualized by ethidium bromide staining. To authenticate the fidelity of PCR amplification, PCR products were subjected to direct sequencing using an automated ABI377 DNA sequencer (Applied Biosystems, Foster City, CA).
Parallel PCR experiments were also conducted for ␤-actin as a system control. The primer pairs used were ATCATGTTTGAGACCT-TCAACAC and caggaggagcaatgatcttg (GenBank accession nos. M10278 and 5016088), and PCR was carried out as described above.

Trans-activation of the CD28 ␣␤-INR Is Inhibited by Mutations in
Site ␣-The CD28 INR consists of two contiguous but noncompeting sequence motifs, ␣ and ␤, which have no homology with the consensus INR and other regulatory elements (10,24,25). These sequences function as a unit; neither ␣ nor ␤ alone has the ability to activate transcription of INR-driven DNA templates (18). Moreover, mutations in either motif result in the complete loss of motif-specific DNA⅐protein complex formation and effectively inactivate the CD28 gene minimal promoter as determined by reporter gene bioassays (10). To ascertain further the relative contribution of either site ␣ or site ␤ in the activity of the ␣␤-INR, the previously described mutated sequences M3, M4, M9, and M10 were introduced in DNA templates and used in transcription assays. As shown in Fig. 1, all of these mutants blocked INR-driven transcriptional activity, which was consistent with the results of reporter assays (10). Clearly, mutations in either ␣ or ␤ motif were sufficient to inactivate INR, further supporting the idea of a structurally bipartite but functionally singular regulatory element (18).
We have reported that the ␣␤-INR is inoperative in CD28 null T cells because of the coordinate lack of ␣and ␤-specific transcription factors (18). Whether such cells were in vivo derived, such as those isolated from patients with rheumatoid arthritis or from healthy elderly donors (18,20), or were generated in vitro from a CD28 ϩ precursor (38), a CD28 null T cell phenotype was directly correlated with the complete lack of ␣-/␤-bound complexes. In the present work, we identified component proteins of the transcription complex which specifically bind site ␣ and assessed their role in the activity of the ␣␤-INR.
Purification of CD28 INR Site ␣-Binding Proteins-To isolate the CD28 INR-specific proteins, nuclear extracts from Jurkat, a prototypical CD28 ϩ T cell line (18), were cleared in a series of agarose columns, beginning with immobilized DNA, followed by heparin and strepavidin. The flow-through from the last adsorption column was subsequently poured into an affinity column consisting of a double-stranded oligonucleotide, corresponding to the site ␣ sequence of the CD28 INR (10), immobilized in a biotin-strepavidin-agarose matrix. DNAbound proteins were eluted by a high salt solution. Retention of the site ␣-specific complexes during column chromatography was monitored by EMSA. As shown in Fig. 2A, this chromatographic strategy enabled the efficient retention and subsequent affinity purification of the relevant DNA-binding proteins. Because the functional CD28 INR consists of two distinct protein binding subsites, ␣ and ␤ (10, 18), DNA binding specificity of the eluted proteins was also confirmed by reciprocal competitive EMSAs using ␣ and ␤ oligonucleotide probes (data not shown). Such assays showed no cross-reactivity between the probes as expected.
In empirical experiments, incubation of precleared extracts with excess amounts of soluble wild type site ␣ oligonucleotide sequences prior to the affinity column purification resulted in a lack of column-bound proteins as confirmed by lack of signals in EMSA and in silver stained SDS-polyacrylamide gels (data not shown).
Chromatographic purification of the CD28 INR site ␣-binding proteins was carried out in three experiments with different batches of nuclear extracts prepared from 75-liter Jurkat cell cultures each. The average yield of DNA affinity columnisolated site ␣-binding proteins was 0.03% (Ϸ130 g) of the starting dialyzed nuclear extracts (Ϸ450 mg).
Similar experiments were conducted using affinity columns consisting of the M3 mutated variant of site ␣. This mutant was chosen because it was the strongest inhibitor of CD28 promoter activity in reporter gene bioassays but does not compete with wild type site ␣ sequences in DNA binding experiments (10). As depicted in Fig. 2B, DNA affinity chromatography experiments using M3 sequences showed a complete lack of proteins that specifically recognize site ␣. The high salt wash from these M3 columns did not contain significant amounts of protein as determined by spectrophotometric assays, silver staining of SDS-PAGE gels (data not shown), or by MALDI-MS (below).
Identification of Site ␣-Binding Proteins-Because the analyte amounts from the affinity column ( Fig. 2A) were limited, we opted to use MS and MS/MS rather than Edman sequencing to identify the bound proteins. Initially, the eluates were subjected to MALDI-TOF-MS to determine the variety and molecular mass of the constituent components. This analysis invariably afforded two distinct protonated molecular ions (MH ϩ ) at Ϸ75 kDa and Ϸ45 kDa (data not shown). Similar MALDI-MS assays of the affinity column flow-through prior to nuclear extract binding or of washes after the salt elution of the affinity column-bound proteins showed no detectable protein ions. Additionally, salt elution from affinity columns of the M3 variant of site ␣ (Fig. 2B) did not yield any column-bound protein as determined by MALDI-MS.
Subsequently, the wild type site ␣ affinity column eluates containing the 75-and 45-kDa components were individually digested with trypsin, preconcentrated on a membrane cartridge, and analyzed by nLC-MS/MS as described elsewhere (26). In the case of the 75-kDa protein, two product ion spectra are shown in Fig. 3A. Product ions from tryptic peptides at m/z 781.47 (Mϩ2H) and m/z 734.10 (Mϩ3H) revealed clear sequence ions of GFGFVDFNSEEDAK and SEDTTEETLKESFD, respectively. Interrogation of the protein data base showed that such peptides were from nucleolin. Similarly, the product ion spectra from the tryptic digest of the 45-kDa protein (Fig. 3B) revealed partial sequence data of VESIELPMDNK (m/z 811.80 (Mϩ2H) 2ϩ ) and FGEVVD (m/z 584.57 (Mϩ3H) 3ϩ ). Subsequent data base search returned that these peptides were from hnRNP-D0.
Results showed that peptide fingerprints of nucleolin and hnRNP-D0 appeared to be the most common, comprising about 70% of the total number of peptides analyzed. These nucleolin and hnRNP-D0 peptide fingerprints were consistently found in three affinity column eluates analyzed independently.
Nucleolin and hnRNP-D0A Are Components of the CD28 INR Site ␣-Bound Complex-To verify that nucleolin, also referred to as C23 nucleolar phosphoprotein (27), and hnRNP-D0 are components of site ␣-binding complexes, EMSAs were performed in the presence of specific antibody. The monoclonal antibody to nucleolin, MS3 (27), was tested. Results showed that addition of this antibody in the DNA binding reactions did not result in band supershifts but was found to inhibit DNA⅐protein complex formation. As shown in Fig. 4, band shifts of site ␣ binding reactions in the presence of MS3 were reduced markedly compared with those containing an IgG isotype control. Regardless of whether the antibodies were added before or after the addition of the oligonucleotide probe to the reactions, MS3 was found to inhibit specific DNA⅐protein complex formation. These results were reproducible in five independent experiments with two dilutions of the antibody as indicated.
Similar EMSAs were also conducted in the presence of anti-hnRNP-D0 antibodies. In these assays, four antisera were examined. These antisera were generated against peptides corresponding to either exon 2 or exon 7, which determines the isoforms of hnRNP-D0 (30, 37, 39) as illustrated in Fig. 5A. These exon-specific antisera have been described previously (28,29). On the one hand, the P1 and P2 antisera were exon 2 region-specific; P1 was specific for exon 2 itself, whereas P2 was directed to a sequence immediately flanking 3Ј of exon 2. On the other hand, the P3 and P4 antisera were exon 7-specific; P4 specificity was exon 7 itself, and P3 specificity was the junction of exons 6 and 7.
As in EMSAs with anti-nucleolin antibodies, the addition of anti-hnRNP-D0 rabbit antisera to the binding reactions did not result in band supershifts. As shown in Fig. 5B, however, antisera to exon 7 (P3, P4), but not exon 2 region (P1, P2), inhibited site ␣-protein complex formation. This suggested that the A isoform of hnRNP-D0, which contains exon 7 but not exon 2, was a component of the site ␣-binding complexes. As expected, the addition of preimmune rabbit serum to the binding reactions had no effect on site ␣ binding activities. These results were reproducible in six independent experiments.
Whether the antisera were added before or after the oligonucleotide probe, only the antisera to exon 7 were found to inhibit site ␣ binding activity.
EMSAs were also conducted with different combinations of anti-nucleolin and anti-hnRNP-D0 antibodies. In these experiments, however, the presence of both antibodies did not totally abrogate site ␣ binding activities (data not shown). Although MS3 or P3 or P4 alone was consistently found to reduce specific

FIG. 4. Site ␣ binding activity is inhibited by anti-nucleolin
antibody. An EMSA for site ␣-specific binding activity (as in Fig. 2) was performed in the presence of the anti-nucleolin monoclonal antibody MS3 at the indicated dilution factors (10, 100) or with an IgG isotype control. DNA⅐protein complexes were fractionated by nondenaturing PAGE and visualized by autoradiography. The radiogram shown is representative of five independent experiments. band shift signals markedly, combinations of antibodies did not eliminate the residual band shifts seen in reactions with a single antibody (as depicted in Figs. 4 and 5B).
Nucleolin and hnRNP-D0A Play a Role in the Trans-activation of the CD28 ␣␤-INR-To assess the role of nucleolin and hnRNP-D0A in the activity of the ␣␤-INR, transcription assays were conducted using INR-driven DNA templates (18). In these assays, nuclear extracts were initially depleted of nucleolin or hnRNP-D0A with specific antibody and then added to in vitro transcription reactions. Depletion of nucleolin or hnRNP-D0A was verified by Western blotting (data not shown), which showed the absence of the protein in the antibody-adsorbed extracts.
As depicted in Fig. 6A, immunodepletion of nucleolin with the MS3 anti-nucleolin antibody resulted in the significant reduction in the levels of ␣␤-INR-dependent transcription. Immunodepletion of nucleolin from nuclear extracts with MS3 resulted in the pronounced reduction or complete abrogation of specific transcripts of ␣␤-INR DNA templates.
Similarly, in vitro transcription assays with nuclear extracts adsorbed with anti-hnRNP-D0A rabbit antisera showed a marked reduction in CD28 ␣␤-INR-dependent transcription. As shown in Fig. 7A, extracts cleared with P3 or P4 anti-hnRNP-D0 exon 7 antiserum (refer to Fig. 5) yielded significantly lower amounts of specific transcripts compared with control extracts or those cleared with preimmune serum. The levels of antisera-dependent reduction of transcripts were equivalent between P3 and P4. Transcription assays with nu-clear extracts cleared with P1 anti-hnRNP-D0 exon 2 antiserum (Fig. 7B) had no effect on the transcriptional activity of the DNA templates.
The role of nucleolin and hnRNP-D0A in the trans-activation

FIG. 5. Site ␣ binding activity is inhibited by antibodies to the A isoform of hnRNP-D0.
A, the isoforms of hnRNP-D0 are identified by the presence or absence of exon 2 or exon 7 (28,29,36) as indicated. RRM, RNA binding motif. B, an EMSA (as in Fig. 2) was performed in the presence of a 1/100 dilution of rabbit antisera to hnRNP-D0 or a preimmune serum control (Pre). The antisera were specific to either exon 2 (P1), the 3Ј-flank of exon 2 (P2), or exon 7 (P3, P4), as indicated (29). The radiogram shown is representative of six independent experiments.
FIG. 6. Trans-activation of CD28 ␣␤-INR-driven DNA templates requires nucleolin. A, nuclear extracts were adsorbed with anti-nucleolin antibody (MS3) or IgG isotype control and used in transcription assays with CD28 ␣␤-INR-driven DNA templates (as in Fig.  1). As a system control, similar assays were conducted using similar DNA templates under the control of the wild type (wt) or mutated variant (mt) of the TdT INR (18,22). The radiogram shown is representative of two experiments, which consisted of three separate reactions for each of the antibody-adsorbed nuclear extracts. B, similar transcription assays of CD28 ␣␤-INR-driven templates were conducted using nucleolin-depleted (as in A) and nucleolin-reconstituted nuclear extracts. Reconstitution was achieved by the addition of proteins eluted from anti-nucleolin (MS3) matrices to the transcription reactions. The radiogram shown is representative of two experiments.  (Figs. 6B and 7B). Although immunodepletion of these proteins abrogated transcription of INR-driven templates, addition of the immunoprecipitates back into the transcription reactions resulted in the reconstitution of transcriptional activity.
The role of nucleolin and hnRNP-D0A as transcriptional activators was specific for ␣␤-INR but not for classical INRs such as that of TdT (34,40). Results of these experiments (Figs. 6A and 7A) also showed that the immunodepletion of either nucleolin or hnRNP-D0A did not affect the transcriptional activities of TdT INR-driven DNA templates. As expected, DNA templates containing the mutated variant of TdT-INR showed negligible amounts of transcripts regardless of the nuclear extract used in the transcription assays.
CD28 INR Site ␣-Specific Complexes Are Distinct from Other Nucleolin-and hnRNP-D0-containing DNA⅐Protein Complexes-Nucleolin and the B isoform of hnRNP-D0 have been reported previously to bind DNA regulatory sequences. In particular, nucleolin⅐hnRNP-D0B has been found to comprise a B cell-specific transcription factor LR1, which specifically binds to Ig switch region sequences (30,31). Additionally, hnRNP-D0B, by itself, has also been reported to bind an enhancer element in the CR2 gene promoter in B cells (28,32). Hence, we examined whether these sequences had overlapping proteinbinding activities with the CD28 INR site ␣ sequence (10,20).
Competitive EMSAs were therefore performed. As expected, excess amounts of unlabeled site ␣ sequences effectively outcompeted the radiolabeled site ␣ oligonucleotide probes as shown in Fig. 8A. In contrast, neither the M3 variant of site ␣, the LR1-specific Ig switch region sequence, nor the CR2 sequence was an effective competitor of the site ␣ binding probes. At 300 molar excess of each competitor, site ␣-bound complex formation was unperturbed.
Consistent with the reported B cell-specific binding activities of LR1 (30,31) and CR2 (28,32), DNA binding assays using these sequences as binding probes for Jurkat T cell nuclear extracts did not demonstrate significant protein binding activities as shown in Fig. 8B. As expected, the M3 mutated variant of site ␣ also lacked protein binding activity.
To assess further that DNA⅐protein complex formation with site ␣ sequences is distinct from that reported for LR1 and CR2 sequences (28,30,31), transcription assays were conducted using Jurkat T cell nuclear extracts that were preincubated with these sequences. As shown in Fig. 8C, incubation of extracts in excess amounts of synthetic double-stranded LR1 or CR2 oligonucleotides did not affect the transcription of CD28 ␣␤-INR-driven DNA templates. In contrast, incubation of the extracts in wild type site ␣ oligonucleotides effectively blocked ␣␤-INR-dependent transcription. As expected, the M3 mutant variant of site ␣ did not block transcription. The transcriptional activity of TdT INR-driven templates were unaffected by the preincubation of the extracts in the site ␣, LR1, or CR2 sequences.
Because nucleolin and hnRNP-D0 proteins are known to bind RNA sequences (36,37,41,42), experiments were conducted to examine whether site ␣ binding activities can be achieved with single-stranded DNA sequences. As shown in Fig. 9, neither sense nor antisense site ␣ oligonucleotides showed significant protein binding activities. This was in marked contrast with high levels of DNA⅐protein complexes found with double-stranded site ␣ sequences.
Preferential Formation of Nucleolin⅐hnRNP-D0A⅐Site ␣ Complexes in CD28 ϩ , but Not CD28 null , T Cells-To examine whether the differential expression of site ␣-binding complexes between CD28 ϩ and CD28 null T cells (18,20) could be attrib-uted to the presence or absence of nucleolin and hnRNP-D0A, Western blotting experiments were conducted. As shown in Fig. 10A, nucleolin was found in the nuclear extracts of both CD28 ϩ and CD28 null T cells. The relative levels of expression of nucleolin, as detected by the MS3 antibody, were equivalent in at least two transformed T cell lines (Jurkat and HUT78), two primary T cell lines (H28P and H28N), and two T cell clones (PL65 and K2) examined. Regardless of the CD28 phenotype of the cell, nuclear nucleolin was detected as a single protein band of Ϸ105 kDa. Consistent with previous reports (41,42), this apparent molecular mass of nucleolin was significantly larger than the predicted size of 76.3 kDa based on the amino acid sequence. Such an increase in the molecular mass had been attributed to post-transcriptional modification of the protein (43). Parenthetically, the MS3 antibody had been shown previously to recognize nucleolin with apparent molecular masses between 100 and 120 kDa (27).
Like nucleolin, hnRNP-D0A was expressed in all of the T FIG. 8. DNA sequence motifs recognized by hnRNP-D0B⅐nucleolin complexes do not inhibit site ␣ binding activity. A, EMSA (as in Fig. 2) was performed using CD28-INR site ␣ sequences as the binding probe in the presence of excess amounts of unlabeled competitor double-stranded oligonucleotides as indicated. Such competitors were the Ig switch region sequence recognized by LR1 (B cellspecific transcription factor comprising a nucleolin⅐hnRNP-D0B complex) (30,31), the hnRNP-D0B binding motif in the CR2 gene (28,29), or the M3 mutant variant of the CD28-INR site ␣ (10). The radiogram shown is representative of three independent experiments. B, EMSA was also conducted using site ␣, M3, LR1, and CR2 sequences as binding probes for Jurkat T cell nuclear extracts. The radiogram shown is representative of three experiments. C, nuclear extracts were incubated with a 100 molar excess of double-stranded site ␣, M3, LR1, or CR2 oligonucleotides and subsequently used in transcription assays (as in Fig. 1) with CD28 ␣␤-INR-driven DNA templates. As a system control, similar assays were conducted with templates under the control of either the wild type (wt) or mutated variant (mt) of the TdT INR. The radiogram show is representative of two experiments. cells examined as shown in Fig. 10A. Regardless of the expression of CD28, hnRNP-D0A, as detected by the P4 antiserum (29), was found as a single protein band of Ϸ45 kDa, which corresponded with the predicted molecular mass of hnRNP-D0A (29,36). The relative levels of its expression were equivalent among all of the cells examined.
The pattern of hnRNP-D0A expression among the different T cells examined was recapitulated in RT-PCR assays as depicted in Fig. 10B. There was a relative abundance of transcripts containing exon 7 sequences, which were characteristic of hnRNP-D0A (37). Direct DNA sequencing (data not shown) confirmed the authenticity of the PCR-amplified products.
Although nucleolin and hnRNP-D0A were expressed in both CD28 ϩ and CD28 null T cells, a complex of these proteins that recognized the CD28 INR site ␣ was found only in CD28 ϩ cells. Using a combination of modified EMSA using biotinylated binding probes in concert with Western blotting of DNA⅐protein complexes isolated by strepavidin-agarose pulldown assay, nucleolin was found to be indeed a component of the site ␣-bound complex as shown in Fig. 11A. With nuclear extracts from CD28 ϩ Jurkat T cells, this experimental strategy showed a single protein band recognized by the anti-nucleolin antibody MS3 both in the nuclear extract and in the isolated site ␣-bound complex. A residual amount of MS3-reactive proteins was also found in the supernatants of the EMSA binding reaction from which the site ␣-bound complexes were isolated. In contrast, similar experiments using nuclear extracts from CD28 null HUT78 T cells showed MS3-reactive proteins only in the whole extract and in the supernatant of the EMSA binding reaction.
From the identical samples of whole extracts, the site ␣-bound complexes, and the supernatants of the DNA binding reactions, the presence of hnRNP-D0A was also examined. As depicted in Fig. 11B, samples from experiments using Jurkat nuclear extracts showed a single protein band recognized by the P4 antiserum to hnRNP-D0A. Similar samples from experiments with HUT78 nuclear extracts also showed the presence of P4-reactive proteins in the whole extracts and in the supernatants of the DNA binding reactions, but not in the isolated site ␣-bound complex.
Because nucleolin and hnRNP-D0A belong to a superfamily of structurally and functionally homologous proteins (36,41), the identical samples were also analyzed for the presence on hnRNP-A1, the most abundant of the hnRNP proteins (36). Using the monoclonal 4B10 antibody specific for hnRNP-A1 (35), immunoblotting experiments showed the relative abundance of 4B10-reactive proteins in the whole extracts of Jurkat and HUT78 as depicted in Fig. 11C. Such 4B10-reactive proteins were also present in the supernatants of the DNA binding reactions. However, there was a complete lack of 4B10-reactive proteins in the isolated site ␣-bound complexes, particularly from those samples isolated from binding reactions with Jurkat extracts, which contained nucleolin⅐hnRNP-D0A complexes (Fig. 11, A and B). DISCUSSION The present work shows that CD28 ␣␤-INR function is regulated, at least in part, by a protein complex that includes nucleolin and hnRNP-D0A. Because these molecules are ubiquitous mammalian proteins (36,43), their binding specificity for the 5Ј-site ␣ sequence of the INR of CD28 is rather surprising and impressive. The finding that nucleolin and hnRNP-D0A peptide ions (Fig. 3) comprise 70% of the total peptides analyzed indicates that these proteins are key components of the site ␣-specific complex. Peptide fingerprints of these proteins were found reproducibly in three independent DNA affinity chromatography-MS/MS experiments. Such peptides were  (27), and hnRNP-D0A was detected by the P4 antiserum (29). Data shown are representative of three experiments. B, total RNA was isolated from CD28 ϩ and CD28 null T cells as indicated and subjected to RT-PCR experiments using amplification primers specific for exon 7 of hnRNP-D0A. Similar experiments were also conducted using primers for ␤-actin, a housekeeping gene. Exon 7 and ␤-actin PCR products were fractionated in 1.5% and 0.8% agarose gels, respectively, and visualized by ethidium bromide staining. Data shown are representative of three experiments. absent from analytes obtained from affinity matrices comprising mutated site ␣ sequences (data not shown).
The notion that nucleolin⅐hnRNP-D0A complexes are site ␣-specific is supported further by several observations. First, specific antibodies to either protein perturb DNA binding (Figs. 4 and 5). Second, isolated site ␣-bound complexes are immunoreactive to both anti-nucleolin and anti-hnRNP-D0A antibodies (Fig. 11, A and B). More importantly, these DNA⅐protein complexes do not contain other ubiquitous proteins such as hnRNP-A1 (Fig. 11C), which are structurally and functionally homologous to both nucleolin and hnRNP-D0A (35,36). And third, immunodepletion of either protein in transcription reactions effectively blocks the production of transcripts of ␣␤-INRdriven DNA templates (Figs. 6A and 7A). Additionally, reconstitution of ␣␤-INR-driven transcription by the addition of specific immunoprecipitates (Figs. 6B and 7B) provides evidence that nucleolin and hnRNP-D0A play a role in the transactivation of the CD28 INR. Collectively, these findings also provide further evidence that INR activity can be attributed to transcription factors other than the general transcription factors or the components of transcription factor IID (22, 44 -46).
Nucleolin and members of the hnRNP family are known for their RNA binding properties, which accounts for their various roles in RNA metabolism including translation and RNA shuttling between the nucleus and the cytoplasm (36,41). However, there is increasing evidence that they bind DNA and serve as specific transcriptional regulators: hnRNP-K has been found to be a general transcription factor that interacts with transcription factor IID (47), and the B isoform of hnRNP-D0 has been reported to be specific regulator of CR2 gene expression in B cells (32). Interestingly, hnRNP-D0B in complex with nucleolin has been found to comprise a B cell-specific transcription factor that recognizes Ig switch region sequences (30,31). Thus, our finding that nucleolin and the A isoform of hnRNP-D0 form a transcription factor complex (Fig. 11, A and B) that participates in the activation of the CD28 INR (Figs. 6 and 7) provides yet another evidence that these proteins have indeed, broad functions including the regulation of DNA transcription.
Considering that CD28 is a T cell-restricted molecule (2) and that Ig class switching is B cell-restricted (48), the findings that nucleolin and the A and B isoforms of hnRNP-D0 are regulators of gene expression (Figs. 6 and 7; Refs. 30 and 31) point to an emerging biological concept that common proteins also regulate cell-specific functions. Because gene expression involves the cooperative interaction of proteins, it may be argued that the transcriptional activity of ubiquitous nuclear proteins could be a result of a functional conformation of particular proteinprotein interactions. Differences in the combinatorial outcomes of such ubiquitous proteins could account for their cell-specific functions. This view is supported by the finding CD28 INR site ␣ does not cross-compete with the Ig switch sequence LR1 or the CR2 sequence (Fig. 8) despite the fact that these DNA sequences bind very similar protein complexes. Although the LR1 sequence is recognized by nucleolin in complex with the B isoform of hnRNP-D0 (30,31), the site ␣ sequence is specifically recognized by nucleolin in complex with the A isoform of hnRNP-D0 (Figs. 3-8). Nucleolin⅐hnRNP-D0A complex likely recognizes double-stranded site ␣ sequences (Fig. 9), which is in marked contrast to nucleolin⅐hnRNP-D0B binding to unwound single-stranded LR1 sequences (30) or to known RNA binding activities of nucleolin and hnRNP-D (36,37,41).
Nucleolin and hnRNP-D0A are found in both CD28 ϩ and CD28 null T cells as detected by Western blots and in RT-PCR experiments (Fig. 10). However, nucleolin⅐hnRNP-D0A complexes capable of binding site ␣ sequences are found only in CD28 ϩ , but not CD28 null , T cells (Fig. 11), suggesting that activity of the CD28 INR is not attributable to the presence or absence of these proteins but to the formation of a functionally active DNA binding complex. A key issue therefore is the basis for their preferential complex formation with the INR site ␣ sequence in CD28 ϩ T cells (10,18,20). Because nucleolin and hnRNPs are modular proteins whose biological functions are regulated by post-transcriptional modifications (36,37,41,42), it is possible that site ␣ complex formation is dependent on such modifications. We have preliminary data showing that phosphorylation contributes to site ␣ complex formation. 2 In particular, phosphorylation on serines and threonines appears to be more critical to complex formation than phosphorylation of tyrosines. Although the exact positions of phosphorylated amino acid residues on nucleolin and/or hnRNP-D0A remain to be examined, this finding supports the notion that differential combinatorial outcomes of protein-protein interactions may well provide the basis for the differential assembly of a nucleolin⅐hnRNP-D0A complex on the CD28 INR site ␣.
Although the present data point to a key role of nucleolin and hnRNP-D0A in CD28 expression, the role of another protein in the preferential site ␣-complex formation in CD28 ϩ T cells, however, cannot be ruled out. This is indicated by the observation that nucleolin and hnRNP-D0A peptides comprise only 2 A. N. Vallejo, unpublished data.
FIG. 11. Nucleolin⅐hnRNP-D0A⅐site ␣-complex formation is permissible in CD28 ؉ but not in CD28 null T cells. Jurkat and HUT78 nuclear extracts (100 g) were precleared in strepavidin-agarose and added to an EMSA binding reaction containing biotinylated site ␣ sequences as the binding probe. DNA⅐protein complexes were isolated by strepavidin-agarose. An aliquot of the precleared nuclear extracts (lane a), the isolated DNA-bound fraction (lane b), and an aliquot of the remaining supernatant of the binding reaction (lane c) were subjected to SDS-PAGE and Western blotting. The presence of nucleolin (A) was detected by the MS3 antibody (27), and hnRNP-D0A (B) was detected by the P4 antiserum (29). As system control, immunoblotting for the RNAbinding protein hnRNP-A1 (35) was also carried out using the 4B10 antibody (C). Data shown are representative of two experiments. 70% of the fragmentation ions of site ␣-bound complexes (Fig.  3). The possible role of yet undefined protein(s) is also suggested by the observed residual site ␣ binding activities in the presence of inhibitory antibodies (Figs. 4 and 5) and by the low but detectable ␣␤-INR-dependent transcription with nuclear extracts depleted of nucleolin and hnRNP-D0A (Figs. 6 and 7).
Previous studies have indicated that CD28 transcription may also be regulated by Sp1 and Egr-1 proteins, which bind to a so-called GR element situated in the first exon of the gene (49). The GR sequence is likely an enhancer element as evidenced by observations that the pharmacologic agent phytohemagglutinin increases the activity of GR-driven reporter gene constructs. It may be noted that the GR element is 30-bp upstream from the first translation codon (49), whereas the ␣␤-INR is situated 130-bp further upstream in the 5Ј-untranslated region (10) and coincides with the transcription initiation site (50). Unlike the enhancer activity of the GR sequence, the ␣␤-INR is a core promoter element regulating the constitutive transcription of CD28 (18). As shown in the present work (Figs. 1, 6, 7, and 8C), ␣␤-INR clearly functions as an initiator that regulates transcription of a heterologous DNA template even in the absence of GR sequences. Whether or not productive transcription of CD28 in vivo involves the interaction of Sp-1/Egr-1/GR and the nucleolin⅐hnRNP-D0A⅐site ␣ complexes remains to be examined.
In summary, data presented here show that nucleolin and hnRNP-D0A are part of a transcription factor complex that binds to the site ␣ of the CD28 INR and participates in the initiation of specific transcription through the INR core promoter element (Fig. 1). Although these are commonly expressed proteins (Fig.  10), our data support the view that cell-specific functions may not always require cell-restricted gene products. Rather, the conformational outcomes of interactions between ubiquitous proteins could impart their cell-restricted functions. The findings that nucleolin⅐hnRNP-D0A and nucleolin⅐hnRNP-D0B complexes are involved in CD28 transcription in T cells (Figs. 2-9,11) and in Ig class switching in B cells (30,31), respectively, are consistent with this hypothesis. Thus, it may be argued that loss of the functional conformations of protein complexes between ubiquitous proteins could also account for changes in cell phenotype and function. Along these lines, we postulate that the emergence of CD28 null T cells in vivo during the course of aging (9, 10) is associated with post-transcriptional structural changes in the nucleolin and the hnRNP-D0A molecules. Because the level of CD28 expression is tightly coupled to the modulation or loss of site ␣-complexes (20), it will therefore be of interest to examine whether nucleolin and hnRNP-D0A are subject to different molecular modifications in biological conditions known to modulate the cell surface expression of CD28, such as chronic activation (20), chronic inflammation (14,15,38), and replicative senescence (8,20). In light of the recent finding that the high frequencies of CD28 null T cells in vivo are correlated with hyporesponsiveness to vaccination in the elderly (51), such studies may allow the molecular dissection of the pathway(s) leading to T cell senescence (13).