Stereochemical Analysis of the Functional Significance of the Conserved Inverted CCAAT and TATA Elements in the Rat Bone Sialoprotein Gene Promoter*

Basal transcription of the bone sialoprotein gene is mediated by highly conserved inverted CCAAT (ICE; ATTGG) and TATA elements (TTTATA) separated by precisely 21 nucleotides. Here we studied the importance of the relative position and orientation of the CCAAT and TATA elements in the proximal promoter by measuring the transcriptional activity of a series of mutated reporter constructs in transient transfection assays. Whereas inverting the TTTATA (wild type) to a TATAAA (consensus TATA) sequence increased transcription slightly, transcription was reduced when the flanking dinucleotides were also inverted. In contrast, reversing the ATTGG (wild type; ICE) to a CCAAT (RICE) sequence caused a marked reduction in transcription, whereas both transcription and NF-Y binding were progressively increased with the simultaneous inversion of flanking nucleotides (f-RICE-f). Reducing the distance between the ICE and TATA elements produced cyclical changes in transcriptional activity that correlated with progressive alterations in the relative positions of the CCAAT and TATA elements on the face of the DNA helix. Minimal transcription was observed after 5 nucleotides were deleted (equivalent to approximately one half turn of the helix), whereas transcription was fully restored after deleting 10 nucleotides (approximately one full turn of the DNA helix), transcriptional activity being progressively lost with deletions beyond 10 nucleotides. In comparison, when deletions were made with the ICE in the reversed (f-RICE-f) orientation transcriptional activity was progressively lost with no recovery. These results show that, although transcription can still occur when the CCAAT box is reversed and/or displaced relative to the TATA box, the activity is dependent upon the flexibility of the intervening DNA helix needed to align the NF-Y complex on the CCAAT box with preinitiation complex proteins that bind to the TATA box. Thus, the precise location and orientation of the CCAAT element is necessary for optimizing basal transcription of the bone sialoprotein gene.

Regulation of class II nuclear gene promoters involves interaction between transcriptional activators and coactivators and accessory proteins with general transcription factors that form the preinitiation complex (PIC) 2 (1,2). Many polymerase II promoters contain a TATA box (3) through which TFIID binds and directs the assembly of other tran-scription factors and RNAPII to form the PIC required to initiate transcription (4). TFIID consists of the TATA-binding protein (TBP), which binds specifically and with high affinity to the TATA box (5,6), and TBP-associated factors (TAFs), which confer the specificity of TFIID for polymerase II promoters. The TFIID complex also provides the primary anchor for the assembly of the multimeric PIC that recruits RNAPII (1,7). TBP is a saddle-shaped protein that binds to the minor groove of the TATA element and is characteristically positioned ϳ30 base pairs upstream of the transcriptional start site. The directionality of transcription is primarily determined by the linear order of the upstream activator sequence relative to the TATA box rather than by the orientation of these elements (8).
A CCAAT box is frequently found in close proximity to promoter elements, and in many cases, a precise distance is required for optimal transcription (9). CCAAT boxes are generally conserved within the same gene across species in terms of position, orientation, and flanking sequences. They are located, in either forward or reverse orientation, as a single copy element between nts Ϫ60 and Ϫ100 of the transcription start site in 30% of eukaryotic promoters (10). The CCAAT box is a targeting sequence of NF-Y as well as CTF/NF1 (CCAAT transcription factor/nuclear factor 1), C/EBP (CCAAT/enhancer-binding protein), and CDP (CCAAT displacement protein) (11,12). NF-Y can be distinguished from the other CCAAT-binding proteins on the basis of a specific DNA sequence requirement flanking the CCAAT element. NF-Y binding to the CCAAT sequence transactivates the basal gene expression of the PIC complex. NF-Y comprises three subunits (subunits A, B, and C), each having a DNA-binding domain, which are required for CCAAT binding and transactivation. Subunits B and C contain a histone fold motif, through which they interact to form a tight dimer. Subunits A and C also contain conserved Q-rich domains that have a transcriptional activation function (13)(14)(15). Because NF-Y associates with TBP and several TAFs, it may recruit preinitiation complexes through TFIID binding (13,16).
Bone sialoprotein, a member of the SIBLING gene family that function in biological mineralization (17), is selectively expressed in mature osteoblasts (18). Basal transcription of the BSP gene requires functional CCAAT and TATA boxes, which are characteristically inverted and separated by only 21 nucleotides (19 -22). Although an inverted CCAAT element (ICE) is known to be functional and is commonly used for binding NF-Y (10), the inverted TATA box in the BSP gene was the first identified in eukaryotic organisms (19,21). The unusual arrangement of the CCAAT and TATA boxes in BSP therefore provides a unique opportunity to study the importance of the relative orientation and separation of these elements in the basal transcription of a natural gene promoter. These studies show that both the position and orientation of the CCAAT element relative to the TATA boxes are important for optimizing basal transcription of the BSP gene.
Preparation of Rat BSP Promoter Constructs-All of the constructs used in this study are based on Ϫ60 BSP Luc, in which Ϫ60 to ϩ60 of the rat BSP promoter sequence was blunt end-ligated into SmaI site of pGL3-Basic (Promega, Madison, WI). The following constructs were generated using the QuikChange site-directed mutagenesis kit (Stratagene, La Jolla, CA): constructs with a single point-mutated CCAAT box, tTTGG; with a reversed CCAAT box, RICE; with the TATA element inverted, tataaa; with both the wild type ICE and TATA reversed, RICEϩtataaa; with 2 nts on each side of the TATA inverted (from GG to CC) with the TATA box, Rvt-flTATA; and with both the ICE and the flanking nts of TATA reversed, RICEϩRflTATA. To prepare 1) a series of promoter constructs with deletions between the ICE (and mutated ICE) and TATA boxes; 2) a series of RICE with flanking sequences inverted; 3) the reversed ICE with 5 flanking nucleotides, (f5-RICE-f5); 4) the ICE in a 3Ј 3 5Ј orientation with five flanking nucleotides, f5-FlipC-f5; and 5) the RICE in a 3Ј 3 5Ј orientation with five flanking nucleotides f5-FlipRICE-f5, Ϫ60 BSP Luc was first digested with NheI/ XhoI. The 130-bp NheI/XhoI fragment was then further digested with MboII to obtain the 84-bp MboII/XhoI "adaptor" fragment. Oligonucleotides with 5Ј-NheI sites and 3Ј-MboII sites and with those required mutations were ordered from Invitrogen Canada Inc. (Burlington, Canada). The annealed complementary oligonucleotides, the 84-bp adapter, and the NheI/XhoI-digested and dephosphorylated Ϫ60 BSP Luc were then ligated at ϩ16°C overnight. All of the constructs were confirmed by sequencing.
Transcription Assays-ROS 17/2.8 and RBMC-D8 cells were plated on 24-well cell culture plates (0.4 ϫ 10 5 /well) 24 h prior to transfection. The cells (60% confluent) were then transfected using Lipofectamine 2000 (Invitrogen). A total of 1.0 g of DNA was used for transfection, either 1 g of BSP promoter construct or 0.5 g of BSP promoter construct plus 0.5 g of NF-YA WT or NF-YA-DN expression vector (23). A Renilla luciferase expression vector phRL-SV40 (Promega) was used as an internal control (1/1000 of total DNA). The cells were harvested 24 h after transfection. The cell lysates were prepared with the dual luciferase reporter assay system (Promega), and luciferase was analyzed on a Berthold Lumat LB-9501.
Chromatin Immunoprecipitation Assays (ChIP)-The ChIP procedure was adapted from Mao et al. (25). HeLa cells were plated on 15-cm plates 24-h prior to transfection. For each pull-down, 1 ϫ 10 7 cells were used. HeLa cells grown to ϳ60% confluence were transfected with total DNA of 18.6 g/plate (Ϫ60 bsp Luc, tTTGG, RICE, or D10), using 93 l/15-cm plate Lipofectamine 2000 (Invitrogen). The cells were incubated for 44 h and fixed in 1% (v/v) formaldehyde. The extracted chromatin was sonicated (7 ϫ 10 s at 13 W to fragment DNA to an average size of 600 bp) and then immunoprecipitated using antibodies that recognize NF-YA (CBF-B (H-209); Santa Cruz Biotechnology, Santa Cruz, CA) or normal rabbit serum as a negative control. The chromatin immunoprecipitate was de-cross-linked with 0.3 M of NaCl. The bound DNA was PCR-amplified using promoter-specific primers, resolved by agarose gel electrophoresis, and then visualized by UV fluorescence. ChIP-PCR primers (5Ј to 3Ј) bsp_fwd (GTGACCGTGTTTGGCT-GCTGAG) and Luc_rvs (TTGCTCTCCAGCGGTTCCAT) generated a DNA fragment of 254 bp.
Statistical Analysis-The results of transcription assays depicted in the figures are a representative set of experiments performed in triplicate that were repeated at least twice. Significant effects of individual expression vectors were assessed using a paired Student's t test. The degree of significance is indicated by asterisks, where * indicates p Ͻ 0.05, and ** indicates p Ͻ 0.01; or diamonds, where ࡗ indicates p Ͻ 0.005, and ࡗࡗ indicates p Ͻ 0.001.

RESULTS
Conservation of the Mammalian BSP Promoters-Comparison of the proximal promoter sequences of the human, rat, and mouse BSP genes has shown a highly conserved ICE and an inverted TATA box, which are separated by 21 nucleotides (18). In addition, there is high sequence identity in the promoter sequences, particularly in the regions flanking the pentanucleotide CCAAT element. Notably, other SIBLING (small integrin-binding ligand N-linked glycoprotein) family members, rodent osteopontin (OPN) and dentin sialophosphoprotein (DSPP) genes also include an inverted CCAAT box upstream from an atypical TATA box (Fig. 1). Whereas the TATA box in BSP is perfectly inverted relative to the consensus TATAAA sequence, in the TATA boxes of the OPN and DSPP promoters the penultimate thymidine is replaced by an adenosine. In the OPN gene the distance between the ICE and TATA elements is also 21 nts, whereas these elements are separated by 22 nts in the DSPP gene. To study the importance of the orientation and relative positions of the CCAAT and TATA elements, we analyzed the effects of a series of mutations on the transcriptional activity of a short promoter construct pBSPϩ60 to Ϫ60 ligated to a luciferase report gene. The strategy employed in these studies is depicted in Fig. 2. Orientation of the CCAAT and TATA Boxes-To determine the importance of the orientation of the CCAAT and TATA boxes in the basal transcription of the BSP gene, mutations were introduced to reverse the orientation of these elements both separately and together. When the pentanucleotide ICE box alone was reversed (RICE) to a 5Ј 3 3Ј CCAAT sequence, transcription measured in ROS 17/2.8 osteosarcoma cells was markedly reduced but not to the extent observed with a single nucleotide mutation in the ICE box, whereas reversing the TATA resulted in a modest but reproducible increase in transcription (Fig. 2), as found previously (20). When both elements were reversed, transcription was reduced but not to the same extent as reversing the ICE box alone, probably because of the more favorable effect of the change in the TATA box orientation. As anticipated, no transcriptional activity was obtained when either the CCAAT or ICE sequence was placed in a 3Ј 3 5Ј orientation.
Flanking Sequences and Transcription-To determine the importance of the flanking sequences, two nucleotides on each side of the TTTATA element were reversed with the inverted TATA sequence (Fig. 3). This change in flanking nucleotides reduced transcription with the ICE box left intact and more extensively when the orientation of the ICE was reversed (i.e. ICE 3 RICE). In comparison, when the flanking sequence was reversed with the ICE (f-RICE-f), transcription was progressively increased with an increase of flanking nucleotides until 5 nucleotides on each side had been reversed. With 5 flanking nucleotides reversed with the ICE, transcription approached the rate for the control WT promoter. However, reversing additional flanking nucleotides, beyond 5 nucleotides, resulted in decreased transcription (Fig. 3). Notably, the three nucleotides on the 5Ј side and on the 3Ј side in the reversed sequence conform to the consensus sequence determined for NF-Y binding ( Fig. 1). When cells were cotransfected with an expression vector for the NF-YA subunit, luciferase activities were increased proportionately for each construct, indicating an enhancement of the NF-Ymediated transcription.
Thus, although the orientation of the TATA element is not critical for TBP binding, the nature of the flanking nucleotides appear to be important. In contrast to the TATA box, the flanking sequence must be reversed with the ICE sequence to obtain a functional CCAAT box.
Flanking Sequences and EMSA-To determine whether the reversed flanking sequences were important for retaining NF-Y interactions, transcription factor binding was assessed by EMSA using nuclear extracts from ROS 17/2.8 cells (Fig. 4). With control oligonucleotides encompassing the ICE box and its normal flanking sequences (WT) gel shifts identified a strong band that was efficiently supershifted with antibodies to NF-YB and two minor more quickly migrating bands that were unaffected by NF-Y antibodies (26). With the RICE sequence, a new major band was seen together with a diffuse faster migrating band that migrated differently from the original minor bands. However, with only two 5Ј-and 3Ј-flanking nucleotides reversed with the RICE, the EMSA pattern showed the original minor bands and the major band shifted with NF-Y antibodies, albeit in a lower amount than seen with FIGURE 2. Transcription assays of BSP promoter constructs with ICE and TATA elements at different orientations. Schematic diagram of cloning strategies used for preparing promoter constructs RICE, f5-FlipC-f5, and f5-FlipRICE-f5 are shown on the left side. NF-Y is depicted as the trimer of YA, YB, and YC. The 5 nts on each side of the ICE box are shown as a gradient bar with the head pointing to the consensus 3Ј side of CCAAT-flanking sequence. Further deletions of 5 or 10 nts between the ICE and the TATA were constructed based on f5-FlipC-f5, and f5-FlipRICE-f5. Shown on the right side is the transcriptional activity of the 1) wild type, Ϫ60 BSP Luc; and the constructs with the following mutations: 2) ICE with A mutated to T; 3) reversed ICE (RICE); 4) f5-FlipC-f5, where the ICE is flipped with 5 nts on both sides; 5) f5-FlipRICE-f5, where the ICE is reoriented to GGTTA along with 5 nts on both sides; 6) TATA box, reversed to a consensus TATAAA sequence; and 7) ICE and TATA elements both reversed. The data (ϮS.D.) represent one of three experiments, which produced consistent results. Significant decreases compared with the WT are indicated by ** (p Ͻ 0.01).
the WT probe. Further increases in reversed flanking sequence produced progressive increases in NF-Y binding, with maximal binding seen with 10 flanking nucleotides reversed. The progressive increase in NF-Y binding was also evident when shifts were performed with recom-binant NF-Y, confirming that the major band in the nuclear shifts was NF-Y and that maximal binding of NF-Y occurred with 10 flanking nucleotides. Although the increases in NF-Y binding corresponded to increases in transcription with Ͻ5 nucleotides reversed, thereafter tran- scription activity was decreased, whereas NF-Y binding was further increased (Figs. 3 and 4).
Separation of the ICE and TATA Boxes-To investigate the importance of position and orientation of the CCAAT box relative to the inverted TATA box, the distance separating these elements was reduced by introducing nucleotide deletions between the ICE and TATA boxes. As a result transcription was progressively decreased with maximal reduction observed after the deletion of ϳ5 nucleotides (Fig.  5). Thereafter, transcription remained low until after 10 deletions, when transcriptional activity was fully restored and was frequently above that observed with the WT promoter. Beyond 10 deletions, transcription was progressively and completely lost. Notably, the decrease after 12 deletions occurred despite the retention of 5 flanking nucleotides on the 3Ј side of the ICE box, suggesting that the decrease was not due to compromised transcription factor recognition. To demonstrate that NF-Y-mediated transcription was maintained following the deletion of nucleotides, ROS 17/2.8 cells were cotransfected with NF-YA, which increased transcription in all of the deleted constructs (Fig. 5B). The study was also replicated in a normal rat bone marrow cell line, RBMC-D8, in which cotransfection with a NF-Y-DN vector suppressed the transcriptional activity consistent with NF-YA-mediated transcription (Fig. 5C). Notably, the degree of stimulation by NF-YA and suppression by NF-Y-DN was variable, which may reflect the effect of positional differences on relative expression levels.
To confirm that NF-Y binds to the CCAAT box in the undeleted and deleted constructs in vivo, chromatin immunoprecipitation analyses were conducted in HeLa cells, which express high levels of NF-Y. Following transfection with the WT construct or the 10-nucleotide-deleted construct (D10), promoter complexes were immunoprecipitated with antibodies to NF-Y (␣CBF-B) and a promoter sequence encompassing the CCAAT region amplified by PCR. A band corresponding to the input DNA was strongly amplified for the WT and D10 constructs (Fig. 6), whereas only a weak amplification was observed with transfected promoter constructs in which the CCAAT box was either reversed (RICE) or included a single base mutation (tTTGG). These studies show that NF-Y binding in vivo is retained in the deletion construct.
Separation of the RICE and TATA Boxes-To determine the relationship between the orientation of the CCAAT box and its distance from the TATA box, deletions were introduced between the TATA and 5f-RICE-f5, in which transcriptional activity is essentially optimal for the reversed CCAAT element. With the ICE sequence reversed, transcription was also reduced progressively with nucleotide deletions (Fig.  7). However, contrasting the results with the WT promoter, the transcriptional activity was further reduced beyond 5 deletions with minimal transcription observed after 10 deletions. To complete these analyses we also examined transcriptional activity when the ICE box sequences, together with 5 flanking nucleotides on each side were reversed (flipped) in direction from their 5Ј 3 3Ј orientation to a 3Ј 3 5Ј orientation (Fig. 2). In both cases, transcriptional activity was lost, and no effect was observed when intervening nucleotides were deleted (Fig.  7), indicating that the 5Ј 3 3Ј direction of the CCAAT box (whether in the ICE or f5-RICE-f5 orientation) is critical for transcription factor binding.
The results obtained in this series of experiments could be explained according to the stereochemical relationship between the orientation and position of the ICE (or RICE) and the TATA elements, as shown in Fig. 8. In the normal promoter the position and orientation of the ICE box in relation to the TATA box is conducive to transcription. However, although the orientation of the CCAAT-binding transcription factor is retained, its position relative to the TATA box is changed as deletions are introduced; each deletion not only shortening the distance from the TATA box but also changing its relative position on the face of the DNA helix. Thus, after five deletions the ICE box has rotated ϳ180°, and its position is maximally removed from alignment with the TATA, and minimal transcription activity is seen. Although the rate of transcription is compromised, presumably because of the energy required to twist the DNA, there would appear to be sufficient distance to allow the intervening DNA to bend and thereby allow interaction between the CCAAT-binding transcription factor complex and PIC proteins. When the ICE sequence is reversed with its flanking sequence (f5-RICE-f5), the DNA is distorted to allow the RICE-binding transcription factor to "flip" orientation and interact with the PIC complex. This appears to be possible because transcriptional activity is similar to the WT constructs and indeed loss of transcription with five nucleotide deletions is similar to the reduction with the WT promoter construct. However, with further deletions there is insufficient intervening DNA to allow the flip to occur, so recovery of transcription, when the RICE and TATA are realigned at 10 deletions, is not observed.

DISCUSSION
These studies have investigated the importance of the stereochemical relationship between the ICE and inverted TATA elements that are required for basal transcription and whose position and orientation are conserved in a large number of eukaryotic genes, including BSP and SIBLING family members. Although reversing the inverted TATA into a classical TATAAA element resulted in a modest increase in transcription, a similar reversal of the ICE box sequence (RICE) caused a loss of transcriptional activity. However, transcriptional activity could be recovered by reversing the flanking nucleotides with the ICE. By deleting nucleotides to reduce the separation of the ICE and TATA boxes, transcription was progressively reduced and subsequently recovered, reaching maximal transcription after the deletion of 10 nts, corresponding to a single turn of the DNA helix. These results suggest that the change in relative positions of the ICE and TATA boxes on the face of the helix compromises interactions between transcription factors that associate with these elements, but that optimal interactions can be achieved with 11 nucleotides separating the elements. In contrast, when the ICE box was reversed together with its flanking sequence (f5-RICE- f5) there was no recovery of transcriptional activity as nucleotides were deleted. Collectively, these results indicate that a minimal length of DNA is needed to provide the flexibility for the TATA and RICE boxes to realign and orientate into a favorable position for transcription factor interactions.
An inverted TATA box was originally described for the AdIVa2 promoter, which is present downstream of the transcription initiation site (27). However, the inverted TATA sequence in BSP, which overlaps a vitamin D3 response element (28), was the first identified in eukaryotic genes (20). Although an inverted TATA box in synthetic promoters had previously been shown to direct transcription in vitro, the transcription efficiency was markedly reduced (29,30). However, transcriptional activity in the BSP gene is only slightly lower (ϳ10%) than when the sequence is converted into a "consensus" TATAAA sequence (20) (Fig.  2). Although the TBP could bind to the inverted TATA box in the opposite orientation on the lower strand, the structural basis of the apparent directionality of TBP binding remains a subject of much discussion and speculation (6, 31). Studies of the interaction between TBP and the TATA box show that A 3 T and T 3 A substitutions at positions 2, 4, and 6 in the TATAAA sequence can be tolerated, assuming that the inverted TATA box can undergo the same conformational change as observed with the adenovirus major late promoter and yeast CYC1 TATA boxes (3) used in the crystallographic studies of TBP-TATA box interactions (32,33). However, in crystallographic analyses TBP binds in the "normal" orientation, and it has been suggested that the slightly different curvatures of the two halves of the TBP ␤-sheet determine directionality (34). Although Juo et al. (35) have suggested that proline 285 (human TBP numbering) determines the direction of TBP binding through its minor groove interaction with the first T-A base pair, this proposal is not supported by mutational analysis (31). Moreover, in contrast to the crystallographic studies, TBP in solution binds the TATA box in both orientations with only a slight preference for the consensus orientation (36), and there are no unfavorable interactions when TBP binds to the TATA box in the reverse orientation (37).
Notably, the deformation of the TATA element by TBP (32, 33, 38) creates a structurally invariant nucleoprotein complex that serves as the receptor for TFIIB and TFIIA (39), which increase the preference for normal orientation of the TBP (40). Thus, although the polarity of transcription is independent of TATA box orientation (8), TBP probably binds in the normal orientation to the inverted TATA box in BSP because a TA-rich sequence with Ts predominant on the 5Ј end appears to provide the necessary binding and orientation. When the flanking nucleotides are inverted a C is placed at position 7. Although crystallographic studies indicate that this substitution can be accommodated through the formation of a C:G Hoogsteen base pair (3), transcription is reduced (Fig. 4).
That the sequence of the TATA box can function well in either orientation or with various T º A mutations is consistent with the ability of TBP to recognize variants of nucleotides that form the minor groove of the TATA element, which is believed to be important in preventing potentially lethal mutations without perturbing subsequent steps in PIC assembly (35). However, this conflicts with the frequent conservation of gene specific TATA box sequences, such as the inverted sequence in BSP. Attempts to explain this paradox include the suggestion that there are variant forms of TFIID that interact specifically with different TATA boxes (41) or that the TFIID may assume a characteristic conformation on binding to a particular TATA sequence to provide optimal interactions with upstream enhancers. Alternatively, some TATA boxes may interact with sequence-specific factors other than TFIID, which synergize with TFIID to activate transcription or to repress transcription in nonpermissive cell types (3). This could be important for the cell-and differentiation stage-specific transcription of the BSP gene in bone formation (18).
That immediate upstream activators may be involved in the directionality of TBP binding and transcription is indicated from the topographical analysis of a synthetic promoter in which an ATTTGCAT octamer was found to direct transcriptional polarity through a downstream TATA complex (8). More recently, the Gal4-VP16 and Gal4AH activators were shown to enhance the orientation and axial specificity of the TBP/TATA complex (40). Because the CCAAT-binding protein NF-Y has been shown to be essential for the recruitment of general transcription factors and RNAPII to the core promoter of the ␥-globulin (42), OPN and ODF/RANKL genes, as well as to the TATA-less E2F1 gene promoter (43), it is conceivable that directionality of transcription could be determined by the proximal CCAAT element that is present in ϳ30% of eukaryotic genes (12). Accordingly, the orientation of the TATA sequence would be irrelevant for basal transcription.   The optimal transcription observed in the BSP promoter after 10 nucleotides were deleted suggests that, given the limited space remaining for NF-Y and TBP binding, NF-Y could be interacting directly with TBP and or PIC components. That NF-Y-mediated transcription is retained following the nucleotide deletions was confirmed by the positive and negative effects of NF-YA and NF-YA-DN expression vectors, respectively, in transcription assays and also by ChIP assays (Figs. 5 and  6). Thus, NF-YC/NF-YB has been shown to interact in vitro with TBP but not with a preformed TBP-TATA element (13,44), which would be in agreement with a role in recruitment of the TBP (16). The NF-YB and NF-YC subunits include histone fold motifs required for dimerization to form a structure that is closely related to H2A/H2B and particularly to the NC2a/NC2b dimers (9). Thus, NF-Y subunits could activate transcription by directly interacting with TBP or by blocking the binding of the repressor NC2 to TBP (45). Alternatively, the NF-Y subunits may promote transactivation by interacting with complementary histone fold structures of TAFs, which are present in the TFIID complex.
In contrast to the TATA box, reversing the ICE box resulted in the abrogation of transcriptional activity, whereas inclusion of flanking nucleotides recovered both NF-Y binding and transcription. However, with the inclusion of more than 8 flanking nucleotides, transcription was reduced despite strong NF-Y binding. Previous studies have demonstrated the importance of the CCAAT flanking sequence and an extended consensus sequence that includes four 5Ј-and three 3Ј-flanking nucleotides identified (46). That the corresponding nucleotides associated with the ICE box in BSP conform to the consensus sequence is consistent with the importance of the flanking nucleotides in the RICE sequence. Further, that NF-Y binding is increased with the inclusion of additional flanking sequence indicates the importance of nucleotides beyond the consensus flanking sequence, which might be anticipated given that the trimeric NF-Y complex spans 24 -26 bp (47,48). Although the decrease in transcriptional activity with the inclusion of the extended flanking nucleotides could be attributed to interference with downstream binding of general transcription factors forming the PIC complex, this seems unlikely given the increased transcriptional activity observed when 10 nts were deleted between the CCAAT and TATA boxes. Because the flanking sequences may be required for the proper distortion of the DNA required for transcriptional activation (9), it appears more likely that these structural changes may be compromised by the altered orientation of the ICE box.
Because transcriptional activity is still evident when the distance between the ICE and TATA boxes are reduced suggests that the DNA can distort to allow interaction between the transcription factors or accessory proteins that bind or associate with these elements. Previous studies have reported that transcriptional efficiency is unchanged when the distance between an upstream "octamer" activator and the TATA box was reduced from 24 to 5 nucleotides (8). In these studies the spacing between the upstream activator and TATA box was shown to be less constrained than between the TATA box and the transcription initiator (Inr) site, which would argue against a possible downstream repositioning of the PIC complex as the distance between the ICE and TATA boxes are reduced. However, although the transcriptional activity reported by Xu et al. (8) was unchanged until 3 nts separated the two elements, the cyclical changes we observed in transcriptional activity in relation to the positioning of the ICE and TATA boxes are consistent with maximal expression obtained when two PEA3/EBS, Ets-binding sites, driving expression of a reporter gene functioning as a minimal transcriptional initiator element, were located on the same face of the DNA helix (49). The two sites could be separated by Ͻ2-3 helical turns but, as we found, transcription was markedly reduced when the ele-ments were displaced by half a helical turn. The results of these studies are also consistent with the increased energy required to change the DNA structure to allow interactions between proteins in the CCAAT and TATA complexes as they are progressively displaced on the DNA helix (Fig. 8). A deletion strategy similar to the one that we have used has shown that activation by CRP (cAMP receptor protein) of Escherichia coli is strictly dependent upon the helical phase between the CRP and RNA polymerase-binding sites (50), whereas the importance of proper helical phasing that we have observed with the ICE and TATA boxes has been shown for two operators involved in the repression of the gal promoters by GalR and HU (51).
That changes in DNA structure can be altered to accommodate transcription factor interactions with the PIC complex has been demonstrated by Dion and Coulombe (52), who used site-specific protein-DNA photo-cross-linking to show that the transcriptional activator GAL4-VP16 interacts with a TFIIA-TFIIB complex assembled on the TATA element. Although the TBP-TFAII-TFIIB promoter topology was not altered significantly by the interaction with DNA, similar to the requirements of the CCAAT and TATA box interactions in our study, looping and bending was required, and the efficiency of transcription was dependent upon the location and orientation of the GAL4-binding site relative to the TATA box. Notably, in our studies the recovery of transcriptional activity after 5 deletions was slower than the initial loss of transcriptional activity, which may be due to the additional clockwise twisting required for realigning the CCAAT and TATA elements or the greater energy required to twist the DNA in an counterclockwise direction.
The ability of the DNA to undergo marked structural changes to accommodate interactions between transcription factors binding to the ICE and TATA boxes is conceivable given the effects of transcription factor binding that has been shown from the crystal structure determination of TBP-DNA (32,33) and from circular permutation analysis of NF-Y⅐DNA complexes (53,54). On binding TBP a sharp 80°bend is induced on the 5Ј side of the TATA box, and the minor groove is dramatically widened (32,33). Similarly, the trimeric NF-Y complex binds to both the major and minor grooves with high specificity and avidity (47) causing the DNA to bend by 60 -80°. In both cases the structural changes imposed on the DNA are thought to be a prerequisite for transcriptional activation. However, it is apparent from our transcriptional analyses that the limits of DNA bending are exceeded as the distance between the f5-RICE-f5 and TATA box is reduced (Fig. 7), because this also requires the reversed ICE box to flip over into the orientation of the wild type CCAAT element. Collectively, we believe these studies provide important insights into the stereochemical requirements of CCAAT and TATA elements in the proximal promoters of the BSP gene that are relevant to a broad range of genes in which these elements cooperate in the regulation of basal transcription.