Identification of a family of cAMP response element-binding protein coactivators by genome-scale functional analysis in mammalian cells

Abstract

This report describes an unbiased method for systematically determining gene function in mammalian cells. A total of 20,704 predicted human full-length cDNAs were tested for induction of the IL-8 promoter. A number of genes, including those for cytokines, receptors, adapters, kinases, and transcription factors, were identified that induced the IL-8 promoter through known regulatory sites. Proteins that acted through a cooperative interaction between an AP-1 and an unrecognized cAMP response element (CRE)-like site were also identified. A protein, termed transducer of regulated cAMP response element-binding protein (CREB) (TORC1), was identified that activated expression through the variant CRE and consensus CRE sites. TORC1 potently induced known CREB1 target genes, bound CREB1, and activated expression through a potent transcription activation domain. A functional Drosophila TORC gene was also identified. Thus, TORCs represent a family of highly conserved CREB coactivators that may control the potency and specificity of CRE-mediated responses.

The completion of draft mouse and human genome sequences has underscored the need for a systematic approach to analyzing mammalian gene function. The human and mouse genomes contain ≈30,000 genes (1–4). However, ≈28% of the predicted human and mouse protein coding genes have no known functional domain or functional classification. The development of large collections of characterized full-length human cDNAs, like that of the Mammalian Gene Collection (http://mgc.nci.nih.gov) (5), and the ability to construct gene specific knockdown reagents using RNA interference (short interfering RNA) (6), should provide a means to systematically assess gene function by reiteratively testing these reagents in cell-based assays. As a test of this approach, a collection of 20,704 predicted full-length cDNAs was assembled and tested individually for activation of the IL-8 promoter. The IL-8 promoter was used as a model complex biological process because its expression is affected by a variety of signaling pathways.

The IL-8 gene, expressed by a number of cell types, encodes a chemokine that induces neutrophil migration and activation. IL-8 expression is associated with the development of diseases such as asthma, arthritis, and cancer. The promoter contains binding sites for the transcription factors AP-1, NF-κB, and C/EBP (for review see ref. 7), which act independently or synergistically in response to extracellular stimuli. Activation by the cytokines IL-1 and tumor necrosis factor α (TNF-α) likely act primarily through NF-κB, whereas other stimuli, including metals, hypoxia, reactive oxygen species, and proteosome inhibition, use a variety of other pathways (8–10).

The high-throughput approach used here identified many components of the AP-1, NF-κB, and C/EBP pathways, and a number of uncharacterized proteins. Computational comparison of screening data and further experiments demonstrated that the IL-8 promoter contained an unrecognized cAMP response element (CRE)-like element that was activated by a protein, termed transducer of regulated cAMP response element-binding protein (CREB) TORC1, which is the founding member of a conserved family of CREB coactivators. Thus, screening of arrayed cDNAs represents an unbiased approach for identification of gene function and elucidation of pathways that regulate complex biological processes.

Materials and Methods

Reporter DNA Constructs. pIL-8-luciferase (Luc) was constructed by insertion of the –1491 to +43 region of the human IL-8 gene into pGL3Basic (Promega). PCR was used to generate a reporter containing the first 160 nucleotides of the IL-8 promoter for introduction of AP-1, C/EBP, and NF-κB site mutations as described (11). The variant CRE (CRE-like) site was mutated to 5′-TCGATCAA-3′. Promoter constructs carrying six copies of the IL-8 CRE-like sequence (pCREL-Luc) or five copies of a member of the S-100 Ca⁺²-binding protein family (CAPL) CRE-like sequence 5′-TGACACAA-3′ (pCREL2-Luc) were prepared by inserting PCR-amplified sequences into pTAL-Luc (Becton Dickinson Biosciences, Clontech, Palo Alto, CA). pCRE-Luc (Stratagene) contains four copies of a consensus CRE.

Full-Length Human cDNA Clones. A number of cDNA libraries were constructed in pCMV-Sport6 vector (Invitrogen) or its derivatives for 5′ end sequencing (Celera Genomics, Rockville, MD). Predicted full-length cDNA clones were isolated and arrayed by using a Q-Bot (Genetix, Boston), and were placed into 384-well plates (Genetix) containing 60 μl of Luria Broth Base (Invitrogen), 8% glycerol, and 100 μg/ml ampicillin. Bacteria were grown in 96-deep-well plates containing 1 ml of Terrific Broth (KD Medical, Columbia, MD), and were then used for production of plasmid DNA by using a BioRobot 8000 with a QIAprep Turbo96 PB protocol (Qiagen, Valencia, CA). For high-throughput screening, the 20,704 cDNA clones were dispensed in 384-well PCR plates at 4 μl per well at 7.5 ng/μl in OPTIMEM (Invitrogen).

Cell Culture and High-Throughput Transfection. Trypsinized HeLa cells were resuspended in complete growth medium at 10⁵ cells per ml and distributed into 51 white 384-well plates (Costar) at 30 μl per well by using a Multidrop 384 (Thermo Labsystems, Vantaa, Finland), and were incubated overnight at 37°C in 5% CO₂. For screening, pIL-8-Luc reporter plasmid was added to OptiMEM I serum-free medium (50 ng of reporter per transfection). FuGENE 6 transfection reagent (Roche Diagnostics, Indianapolis) was used for high-throughput transfections at 3 μl of FuGENE 6 per μg of total DNA. Three microliters of OptiMEM-reporter-FuGENE 6 mix was added per well containing 4 μl of prediluted cDNAs by using a BiomekFX (Beckman Coulter). After 15 min incubation at room temperature, 6 μl of the mix from each well was transferred to a 384-well tissue culture plate and incubated for 48 h. Cells were treated with forskolin (1 μM final) or PMA (1 ng/ml), respectively, for 16 h, before Luc assay for some experiments as described.

Luc and ELISA. Luc and ELISA data represent at least triplicate values and SD, except for those of Fig. 1 B and C. Forty-eight hours posttransfection, firefly Luc activity was measured by using a BrightGlo Luc assay system (Promega), following the manufacturer's protocol. Luminescence was determined with a LUMINOSKAN ascent luminometer (Thermo Labsystems) with a 200- or 400-msec integration time. Results were normalized for transfection efficiency by using a cotransfected Renilla Luc plasmid (pRL-SV40; Promega), unless otherwise indicated. Luc activities were measured by using a DualGlo Luc assay system (Promega), according to the manufacturer's protocols. The results in Fig. 2A are absolute values, because different activators affected the control Luc gene differently. For ELISAs, HeLa cells were cotransfected by using 100 ng of one plasmid, listed on the x axis, and 25 ng of a second plasmid, as indicated in the legend, in 96-well plates (Costar). Secreted IL-8 was measured 72 h posttransfection by using an IL-8 ELISA kit (Sigma). Cells transfected with empty vector and treated with IL-1β (R & D Systems) at 5 ng/ml for 16 h were used as a positive control.

Fig. 1.

High-throughput screening of arrayed full-length cDNAs. (A) Percentage of cDNAs with specific InterPro annotations. (B) Fold induction of pIL-8-Luc after transfection of the 10% most active cDNAs. (C) Fold induction (shown by color) by cDNAs in the primary (IL-8_1) and confirmatory IL-8-Luc (IL-8_2) assays, CRE-Luc, and serum response element-Luc reporter assays. (D) HeLa cells were treated with IL-1β or transfected with an empty vector (CMV), or with relA or MAP3K11 constructs (x axis), individually or in combination with a second construct as shown in the legend. The level of secreted IL-8 protein in the media 48 h posttransfection is indicated.

Fig. 2.

TORC1 activates the IL-8 promoter through a CRE-like site. (A) IL-8 promoter constructs with mutations in the C/EBPβ (δC/EBP), NF-κB (δNF-κB), AP-1 (δAP-1), or CRE-like (δCRE-like) sites were transfected with MAP3K11, relA, or TORC1 as indicated. Firefly Luc levels are shown as arbitrary units (AU). (B) CRE-like and consensus CRE-containing reporters were transfected with the activators shown in the key. pTAL-Luc contains the same promoter as the CRE vectors without any response element. Luc values are shown as AUs after normalization. (C) A dominant mutant of CREB1, KCREB, but not IkBα, blocked TORC1 induction. The pIL-8-Luc reporter was cotransfected with vector (CMV) or TORC1, as indicated in the key, and inhibitor as indicated on the x axis. (D) pIL-8-Luc was cotransfected with either δ59 or an empty vector (CMV), as indicated in the key, with various activators as shown on the x axis. (E) pAP-1(PMA)-Luc was transfected with control vector or δ59 as indicated in the key, and cells were treated with PMA or transfected with MAP3K11 as indicated on the x axis.

Gene Expression Profiling with Affymetrix DNA Microarray Chips. HeLa cells were transfected by using Targefect F1 reagent (Targeting Systems, Santee, CA), according to the manufacturer's instructions. Total RNA was extracted 72 h posttransfection by using TRIzol (Invitrogen) and purified by using a Qiagen RNeasy kit (Qiagen), according to the manufacturer's instructions. RNA was quantified by using Ribogreen (Molecular Probes). Isolated RNA (5–10 μg) was amplified and labeled by using a modified Eberwine protocol (12). Biotin-labeled RNA was hybridized to HG_U95A chips, according to the manufacturer's instructions (Affymetrix, Santa Clara, CA). The scanned images were quantified with Affymetrix GeneChip microarray suite 4 with a target intensity of 500. At least two chips were used for each condition.

Association of TORC1 with CREB1. FLAG-tagged constructs were transfected into HEK293 cells in 100-mm dishes (Falcon) by using FuGENE 6, according to the manufacturer's protocol. Lysates were prepared 40 h after transfection in 800 μl of 10 mM Hepes, pH 7.6/250 mM NaCl/5 mM EDTA/1 mM DTT/0.1% Nonidet P-40, and freshly dissolved protease inhibitors. Immunoprecipitation was carried out by using anti-FLAG-M2-agarose beads (Sigma). Precipitated proteins were separated on SDS/4–20% PAGE (Invitrogen) and transferred to a nitrocellulose membrane (Invitrogen). Western blots were performed by using antibody against CREB1 (Cell Signaling Technology, Beverly, MA) and anti-FLAG M2 monoclonal antibodies (Sigma).

TORC1-Related Proteins. Human TORC2 and 3 (hTORC2 and hTORC3) were identified by searching public and inhouse EST databases. hTORC2 was identified from XP_117201. hTORC3 was identified with FLJ00364. Full-length cDNAs were constructed in pCMV-Sport6. The sequences shown are predicted from clones isolated here. Drosophila TORC (dTORC) was derived from the predicted gene CG6064. A 2.3-kb cDNA encoding dTORC was used to produce pUAS-dTORC. CRE-hsp-Luc contains four copies of a Drosophila CRE, followed by a heat shock 70-kDa protein minimal promoter. Drosophila melanogaster Schneider cells (S2) were transfected in six-well plates (Costar) by using calcium phosphate. A total of 25 μg of DNA was transfected into a six-well dish containing 4 ml of cells (10⁶ cells per ml). Luc assays were performed 48 h later. The UAS-transgenes were activated by cotransfection with the actin promoter-GAL4 plasmid provided by Norbert Perrimon (Department of Genetics, Harvard Medical School, Boston). Values were normalized by using a cotransfected Renilla Luc gene driven by a heat shock promoter. As a negative control S2, cells were cotransfected with CRE-hsp-Luc reporter and empty pUAST vector.

Results

Putative full-length cDNAs (20,704) were tested for the ability to induce the IL-8 promoter on transient transfection in human cells. The cDNA clones were selected, based on the presence of a start codon or 5′ end of a predicted or characterized gene. The 12,905 clones matched RefSeq genes, of which 5,463 were functionally annotated (Fig. 1A). Complete sequencing of >100 clones suggested that ≈70% of the rescued cDNAs contained full-length cDNAs (data not shown). The cDNAs were cotransfected with a firefly Luc gene controlled by the IL-8 promoter (pIL-8-Luc). Sixty-four cDNAs induced the reporter by >5-fold (Fig. 1B). The verified active cDNAs included one to three copies of 28 unique genes. The entire collection was also screened for activation of CRE- or serum response element-driven reporters. Twenty-two unique cDNAs were chosen for further investigation, and the activities of these clones in the primary screens and a secondary IL-8-reporter assay are shown after hierarchical clustering analysis (Fig. 1C). Genes relatively specific for the IL-8 reporter (where the induction of the IL-8 reporter was >5-fold higher than other reporter constructs) included several known inducers of NF-κB represented by relA, a subunit of NF-κB, the TNF receptor TNFSFR1A, TNF-related molecule TNFSF12, RIPK2, TRAF6, ACT1, and the kinase ANKRD3. Genes known to activate AP-1 were also identified, including the JNK-inducing kinases MAP3K11 and MAP3K12, junD, and c-jun (data not shown). C/EBPβ, predicted to bind the C/EBPβ site, was also identified. Several cDNAs were identified whose mechanisms of action were not clear, including two guanine nucleotide exchange factors for the Rho GTPase (Rho-GEFs), p114 and ARHGEF1, C16orf15, thyrotroph embryonic factor 1 (TEF1), fibronectin (FN1), and nuclear receptor family member NR2F2. C16orf15 encodes a proline-rich protein of unknown function (13). TEF1 is a member of the basic leucine zipper transcription factors, which acts directly through a TEF response element (14). FN1 is a matrix glycoprotein highly expressed in injured tissues reported to induce the IL-1 gene through AP-1 (15).

The accumulation of secreted endogenous IL-8 from HeLa cells was measured after transfection with relA and MAP3K11 alone or in combination (Fig. 1D). MAP3K11 and relA induced small increases, but the combination of both induced levels of secreted IL-8 comparable to that observed after treatment with IL-1β, one of the most potent inducers of IL-8 known. Thus, whereas the reporter gene was potently induced by proteins activating a single pathway, activation of multiple signaling pathways synergistically induced expression of the endogenous IL-8 gene.

Surprisingly, a number of CRE-binding proteins CREB1, CRE-BPa, and XBP1, potently activated the IL-8 reporter, suggesting that it contains an unrecognized CRE site. Further, cDNAs encoding C/EBPβ, JunD, and a clone overlapping with KIAA0616 were potent inducers of both the CRE and IL-8 reporters. The CRE-like site was identified with the sequence 5′-TGACATAA-3′ at –69 to –61 of the IL-8 promoter. This sequence contains two changes from a consensus CRE, both of which were reported to be tolerated for binding to either CREB1 or CREB2 (16).

The mechanism of induction by MAP3K11 and KIAA0616 was pursued, because they were the strongest activators of the IL-8 promoter found. No function is known for KIAA0616, except that sequence encoding its first 44 amino acids was recently found translocated onto the Mastermind-like gene, MAML2, in mucoepidermoid carcinoma (MEC) (17). As KIAA0616 and several related genes are shown here to specifically activate CREB-dependent gene expression, (see below) this gene was designated as TORC1.

IL-8 promoters carrying mutations in the CRE-like and other regulatory sites were tested for induction by MAP3K11, TORC1, or relA (Fig. 2 A). Mutation of the C/EBPβ-binding site had no effect. The NF-κB site mutation had little effect on induction by MAP3K11 or TORC1, but eliminated induction by relA. Mutation of the AP-1 site did not significantly affect response to relA, but severely reduced induction by MAP3K11, which was consistent with its known ability to activate the JNK/SAPK pathway and AP-1. This mutation also significantly reduced, but did not eliminate, activation by TORC1.

Mutation of the CRE-like site dramatically decreased induction by both TORC1 and MAP3K11. The ability of these proteins to activate a minimal promoter carrying concatamerized CRE-like sites (pCREL-Luc) was then examined. pCREL-Luc was strongly activated by TORC1, but not by the MAP3K11 or PMA, an AP-1 activator (Fig. 2B). TORC1, but not MAP3K11 or PMA, also potently induced a promoter driven by consensus CRE-sites (pCRE-Luc). In contrast, MAP3K11 potently induced an AP-1 reporter, which was relatively unaffected by TORC1 (data not shown). Thus, TORC1 acts through CREs, whereas MAP3K11 acts as an AP-1 inducer.

The effect of a KCREB1 construct, containing an inactivating mutation in the DNA-binding domain (18), was examined to determine whether TORC1 acted through a CREB-related protein. Coexpression of KCREB with TORC1 significantly reduced induction of the IL-8 promoter (Fig. 2C). In contrast, TORC1 activity was unaffected by cotransfection with I-κBα, which potently inhibits NF-κB. These data suggest that TORC1 acts at least partly through CREB1 or a related protein.

Both the AP-1 and the CRE-like sites were required for activation of the IL-8 promoter by MAP3K11 or TORC1. To determine whether TORC1 participated in this interaction between these sites, a dominant interfering variant of TORC1 was tested for its ability to affect activation. A TORC1 mutant, δ59, missing the first 59 amino acids, was largely inactive for reporter induction, and reduced activation by cotransfected wild-type TORC1 by 70% (Fig. 2D). This mutant greatly inhibited activation of the IL-8-Luc by MAP3K11, but had no effect on activation by relA. These data suggest that TORC1 may specifically regulate activation of the IL-8 promoter through the AP-1 and CRE-like sites. Activation of an AP-1 specific reporter by PMA or MAP3K11 was also blocked by δ59 (Fig. 2E). No effect was seen by over expression of an I-κBα construct (data not shown). These data suggest that TORC1 may interact with an unidentified protein, which is essential for AP-1-mediated transcription, and that TORCs may regulate a subset of AP-1-dependent responses.

HeLa cells were transfected with TORC1, MAP3K11, or relA constructs, or treated with TNF-α and gene expression profiles were compared by using DNA microarrays to determine whether TORC1 specifically induced expression of endogenous CRE-dependent genes. TORC1 transfection induced seven genes by >10-fold (Fig. 3A), including several well known targets of CREB1, namely TSHα (19), phosphoenol pyruvate carboxykinase (PEPCK) (20), crytallin α-B (21), amphiregulin (22), and to a lesser extent, CREM (23). This set of genes was unaffected by MAP3K11, which induced PAI-2, a known AP-1 target gene (24), or relA. Thus, TORC1 specifically induced authentic CREB1 target genes. The endogenous IL-8 gene was modestly induced (2.5- to 5-fold) by each activator tested, which was consistent with a requirement for activation of multiple pathways for efficient induction of the endogenous IL-8 gene. Both relA and TORC1 potently activated expression of the chemokine MIP3α and KIAA0467, suggesting their promoters have both NF-κB and CRE-like sites.

Fig. 3.

TORC1 induces known CREB1 target genes. (A) Fold activation of mRNA levels was measured on Affymetrix U95a chips after TNF-α treatment or transfection of HeLa cells with MAP3K11, p65, or TORC1 constructs. Values represent the averages of two experiments. (B) TORC1 activates expression of a CRE-like sequence (pCREL2-Luc) found in the CAPL and PEPCK promoters. Each reporter, indicated on the x axis, was cotransfected with TORC1 and analyzed for activation compared with vector transfected cells. (C) Cells transfected with the vector indicated on the x axis were treated with forskolin (FSK), cotransfected with CRE-BPa, or both, as indicated.

CAPL, potently induced by TORC1, is not known to contain a CRE site. A sequence similar to the CRE-like site was identified designated CRE-like2, (5′-TGACACAA-3′) at –385 and –392 of the predicted CAPL promoter. The CRE-like2 element, when placed upstream of a minimal promoter, allowed induction by TORC1 (Fig. 3B). To determine whether the TORC1-responsive elements acted as true CREs, their response to forskolin or a CREB-related protein (CRE-BPa) was examined. The IL-8 promoter, CRE-like, and CRE-like2 reporters were modestly activated by forskolin and synergistically induced by both forskolin and CRE-BPa (Fig. 3C). Thus, to date, all TORC1-responsive elements represented true CREs and supported the notion that TORC1 works primarily through CREB1 or other CREB-related proteins that might have higher affinity to variant CRE-sites.

Sequence databases were searched for TORC1-related proteins. Whereas no significant homologies to characterized proteins were found, a number of uncharacterized proteins highly related to TORC1 were discovered. TORC1 orthologs were identified in mice and fugu, which shared 90% and 66% amino acid identity, respectively (data not shown). Two TORC1-related human genes (hTORC2 and hTORC3, which are 32% identical to TORC1) and a single dTORC (which is 20% identical to TORC1), were identified as shown (Fig. 4). The proteins display a highly conserved predicted N-terminal coil–coil domain (hTORC1 residues 8–54) and an invariant sequence matching a protein kinase A (PKA) phosphorylation consensus sequence (RKXS). hTORC2 and hTORC3 cDNAs were isolated and their ability to activate CRE-driven expression was tested. Both TORC1-related genes potently induced expression from the pCRE-Luc and pIL-8-Luc reporters similar to TORC1 (Fig. 5A). A cDNA for a potential Drosophila ortholog, dTORC, was isolated and tested for activity in Drosophila S2 cells. Transfection of the dTORC expression plasmid potently induced a heat shock promoter when linked to Drosophila CREs in S2 cells. (Fig. 5B). Thus, these genes represent a functionally conserved family of proteins, which potently induce CRE-driven gene expression.

Fig. 4.

TORCs are a conserved gene family. Shown is alignment of hTORC- and dTORC-predicted proteins. Conserved amino acids are shaded. Alignments were produced with clustalw.

Fig. 5.

TORC proteins are CREB1 coactivators. (A) Constructs encoding the three human TORC genes (indicated on the x axis) were transfected with either pIL8-Luc or pCRE-Luc, as indicated. Values shown are fold induction, as compared with vector-transfected cells. (B) dTORC expressed in S2 cells induced a reporter carrying Drosophila CREs. Cells were transfected with the indicated reporter and activator as indicated on the x axis. (C) The N-terminal domain of hTORC1 interacts with CREB1. HEK293 cells were transfected with the indicated FLAG-tagged constructs encoding either amino acids 1–170 or 170–650 of TORC1, or human histone deacetylase 1 (HDAC1) as indicated. Protein complexes isolated with anti-FLAG antibody were tested for the presence of CREB1 by Western blot (Upper). The first lane shows CREB1 detected in whole-cell extracts. The same filter was also tested with anti-FLAG M2 antibody (Lower). (D) hTORCs contain potent transcriptional activation domains. HEK293 cells were cotransfected with UAS-Luc reporter plasmid and full-length TORC constructs or GAL4BD-TORC1 (amino acids 300–650), TORC2 (amino acids 296–694), or TORC3 (amino acids 335–635) fusions. Fold induction is relative to transfection with the GAL4-DNA-binding domain vector, pCMV-BD. The reporter was also transfected with a plasmid encoding GAL4-CREB fusion protein alone (GAL4-CREB), or with a PKA catalytic subunit expression construct (GAL4-CREB/PKA) as a positive control.

Several observations suggest that TORCs are CREB1 coactivators. Epitope-tagged TORCs are localized in the nucleus (data not shown). However, their coding regions contain no obvious DNA-binding domain, and experiments failed to demonstrate DNA binding by TORC1 (data not shown). Each protein contained a serine/glutamine-rich domain (hTORC1 residues 289–559) and a negatively charged C terminus similar to transcription activation domains. Regions of TORC1 were tested for association with endogenous CREB1. The N-terminal 170 amino acids of TORC1, containing the highly conserved coil–coil domain essential for TORC activity, was found to be associated with endogenous CREB1 in human 293 cells (Fig. 5C). To determine whether the TORCs contained a transcription activation domain, various regions of the human TORCs were expressed as fusion proteins with the DNA-binding domain of GAL4 (GAL4BD), and tested for induction of a minimal promoter linked to GAL4-binding sites (UAS-GAL4). UAS-GAL4 was potently induced by GAL4BD fusions containing the C-terminal portion of all three human TORCs, but not by the full-length TORC proteins (Fig. 5D). Thus, the TORC proteins represent coactivators that bind CREB1, and, presumably, bring into the complex a potent activation domain. It should be noted that activation by GAL4-TORCs or TORCs was greater than that observed by either activated GAL4-CREB1 or CREB1, respectively. In other experiments, all three TORCs enhanced activation by a GAL4-CREB1 fusion, which was consistent with TORCs acting as CREB1 coactivators (data not shown).

Discussion

CREB1 is one of the best studied inducible eukaryotic transcription factors. CREB1 is induced by phosphorylation by PKA, which allows association with the coactivator CBP. CREB1 regulates a large number of genes controlling cell growth, survival, metabolic control, and memory. How CREB1 and related proteins achieve cell-type- or stimulus-specific induction of subsets of genes is not understood. The evolutionary conservation of TORCs strongly suggests they will have an important role in regulating CREB1-dependent responses. As the TORCs potently activate CREB-dependent gene expression in the absence of extracellular stimuli, TORCs may act as a rheostat for controlling the magnitude of CREB responses, and may provide an intracellular mechanism to control CREB-responsive gene expression during development. The presence of a conserved, potential PKA phosphorylation site in all TORCs suggests that TORCs and CREB1 could be coordinately regulated.

The ability of the human TORCs to activate CRE-like elements weakly responsive to cAMP may be explained in several ways. TORCs may interact with other CREB proteins such as CREB2 or CREB-BPa with alternate DNA-binding specificities. It should be noted that TORCs were also identified in an independent study (25) that demonstrated that TORC1 could not effect a consensus CRE in CREB1-deficient fibroblasts. Alternatively, TORCs may selectively activate a subset of responsive genes by interaction with other transcription factors, which is consistent with the observation that induction of IL-8 by TORC1 required both the AP-1 and CRE-like sites. Elucidation of TORC-mediated specificity should shed light on the selective activation of CRE-containing genes.

As discussed above, a chromosomal translocation between the first 44 amino acids of TORC1 and the Mastermind-like 2 gene (MAML2) was found to create a transforming gene (17). Because the first 59 amino acids of TORC1 are shown here to be critical for recognition of CREs, transformation by the TORC1–MAML2 fusion might result from uncontrolled activation of CREB-responsive genes.

The experiments described here raise the question of the importance of the CRE-like site in regulating IL-8. In this regard, β₂-adrenergic agonists (β₂-AR), which increase intracellular cAMP levels, were shown to induce IL-8 secretion in monocytes and bronchial epithelial cells (26, 27). This fact is particularly important, because the use of β₂-AR agonists as bronchodilators can exacerbate asthma and should be used in conjunction with anti-inflammatory steroids (28). This study suggests β₂-AR may directly induce IL-8 transcription through the CRE-like site.

This article demonstrates the utility of high-throughput screening of curated cDNAs in mammalian cells. Activities were assigned to several previously uncharacterized genes and added to the characterization of several known genes. The utility of reiteratively testing gene sets against many assays and building databases of function was illustrated by the prediction that the IL-8 gene would have a CRE sequence.

Note. While this work was being completed, a similar approach was reported that used a smaller set cDNAs encoding potentially secreted proteins (29). This article and the work presented here suggest that high-throughput screening for function of cDNAs will find wide application in the analysis of complex biological systems.