The CYP2B2 phenobarbital response unit contains an accessory factor element and a putative glucocorticoid response element essential for conferring maximal phenobarbital responsiveness.

Hepatic cytochrome P450s play a critical role in the metabolism of hydrophobic xenobiotics. One of the major unsolved problems in xenobiotic metabolism is the molecular mechanism whereby phenobarbital induces hepatic enzymes, particularly CYP2B1 and CYP2B2 in rat liver. By using primary rat hepatocytes for transfection analyses, we previously identified in the CYP2B2 5'-flank a 163-base pair Sau3AI fragment that confers phenobarbital inducibility on a cat reporter gene and that has the properties of a transcriptional enhancer. Transfection experiments with sub-regions of the Sau3AI fragment now indicate that a central core together with an upstream or downstream accessory element within the fragment can confer phenobarbital responsiveness. One such accessory element, AF1, was identified and localized. DNase I footprinting analysis revealed the presence of a footprint overlapping this AF1 element. It also identified three other major protected regions, two of which are putative recognition sites for known transcription factors. Site-directed mutagenesis indicated that a putative glucocorticoid response element as well as a nuclear factor 1 site and an associated nuclear receptor hexamer half-site are essential for conferring maximal phenobarbital inducibility. Taken together, the results indicate that phenobarbital induction of CYP2B2 requires interactions among multiple regulatory proteins and cis-acting elements constituting a phenobarbital response unit.

Hepatic cytochrome P450s play a critical role in the metabolism of hydrophobic xenobiotics. One of the major unsolved problems in xenobiotic metabolism is the molecular mechanism whereby phenobarbital induces hepatic enzymes, particularly CYP2B1 and CYP2B2 in rat liver. By using primary rat hepatocytes for transfection analyses, we previously identified in the CYP2B2 5-flank a 163-base pair Sau3AI fragment that confers phenobarbital inducibility on a cat reporter gene and that has the properties of a transcriptional enhancer. Transfection experiments with sub-regions of the Sau3AI fragment now indicate that a central core together with an upstream or downstream accessory element within the fragment can confer phenobarbital responsiveness. One such accessory element, AF1, was identified and localized. DNase I footprinting analysis revealed the presence of a footprint overlapping this AF1 element. It also identified three other major protected regions, two of which are putative recognition sites for known transcription factors. Site-directed mutagenesis indicated that a putative glucocorticoid response element as well as a nuclear factor 1 site and an associated nuclear receptor hexamer half-site are essential for conferring maximal phenobarbital inducibility. Taken together, the results indicate that phenobarbital induction of CYP2B2 requires interactions among multiple regulatory proteins and cis-acting elements constituting a phenobarbital response unit.
Many different cytochrome P450s (CYPs) 1 are involved in the hepatic metabolism of a wide variety of xenobiotic substances including drugs, plant metabolites, and chemical carcinogens (1,2). The genes encoding these enzymes are either expressed constitutively or are induced by various chemicals (3). One of the major unsolved problems in the study of the induction of CYP proteins is the molecular mechanism whereby phenobarbital (PB) induces the closely related CYP2B1 and CYP2B2 forms in rat liver (4). Although it has long been known that PB induces CYP2B1 and CYP2B2 mRNAs by increasing transcription of their genes (5,6), details of the transcriptional control of CYP2B1 and CYP2B2 have not been forthcoming. This is largely because, until recently, it was not possible to obtain PB induction of the endogenous CYP2B1 and CYP2B2 genes in cultured cells (4,7,8). We have transfected reporter gene constructs into cultured adult rat hepatocytes and localized, in the CYP2B2 5Ј-flanking region, a 163-base pair (bp) Sau3AI fragment that confers PB inducibility on a cat reporter gene (9). The Sau3AI fragment, which is situated between 2155 and 2317 bp upstream of the CYP2B2 transcription start point, in the vicinity of a liver-specific DNase I-hypersensitive site (10), has the properties of a transcriptional enhancer (9). The capacity of the Sau3AI fragment to confer PB responsiveness on a heterologous promoter has been confirmed in a quite different assay system involving in situ DNA injection into rat liver (11). Furthermore, the homologous region of the 5Ј-flanking region of the PB-inducible mouse Cyp2b10 gene contains a segment 91% identical to the rat CYP2B2 163-bp fragment that also confers PB inducibility on heterologous promoters and possesses the properties of a transcriptional enhancer (12).
The CYP2B2 163-bp Sau3AI fragment contains a functional nuclear factor 1 (NF1) site (9,13) and recognition sites for other sequence-specific DNA binding factors present in rat liver nuclear extracts (9). In the present study, we sought to define the elements within the 163-bp Sau3AI fragment which confer PB responsiveness. Deletion constructs of the Sau3AI fragment, as well as constructs in which putative recognition sites for DNA binding factors had been mutated, were transfected into adult rat hepatocytes, and their effect in conferring PB responsiveness was analyzed. DNase I footprinting experiments were used to identify potential regulatory elements by defining interactions between rat liver nuclear proteins and the Sau3AI fragment. A putative glucocorticoid response element as well as an NF1 site and an associated nuclear receptor hexamer halfsite were found to be essential for conferring maximal phenobarbital inducibility. Taken together, the results indicate that the Sau3AI fragment is a multicomponent enhancer and that multiple regulatory proteins and their cognate recognition sequences within it are critical for obtaining maximal PB responsiveness. Thus, the Sau3AI fragment constitutes a PB response unit (PBRU), analogous to the complex glucocorticoid response unit (GRU) required for glucocorticoid induction of transcription of the rat gene for phosphoenolpyruvate carboxykinase (PEPCK) (14 -17 Isolation and Culture of Primary Hepatocytes, Transfection, PB Treatment, and Chloramphenicol Acetyltransferase (CAT) Assays-The methods for hepatocyte isolation and culture, essentially those of Waxman et al. (18), as well as those for liposome-mediated transfection (19) and PB treatment have been described (9). Plasmids were purified using a plasmid purification kit (Qiagen). CAT activity was assayed by the method of Gorman et al. (20) and, occasionally, by that of Seed and Sheen (21).
Construction of Deletion and Point Mutants-Sequential 5Ј or 3Ј deletions of the 163-bp Sau3AI fragment (Ϫ2317/Ϫ2155) were generated by restriction enzyme or Bal31 digestion (Fig. 1). The starting point for construction of such deletion mutants was the pSa-Sa163 plasmid, obtained by subcloning the 163-bp Sau3AI fragment into the SmaI site of pBluescript KS (Stratagene). The pSa-Sa163 plasmid or similar pBluescript KS derivatives containing specified portions of the CYP2B2 5Ј-flank were used to generate other deletions by polymerase chain reaction-mediated amplification. To obtain Ϫ2257/Ϫ2208 (Fig. 1) the starting pBluescript derivative contained Ϫ2257/Ϫ2012, and the primers were KS (Stratagene) and oligo HX (GATCCACTGTGCCAAG-GTCAGGA Ϫ2226, lower stand; the underlined nucleotides were added for convenience). In a series of amplifications ( Fig. 1), primers 2B2Nco (Ϫ2257 CATGGTGATTTCAGGCA) and SK (Stratagene) were used as follows: to obtain Ϫ2257/Ϫ2207 the starting pBluescript derivative contained Ϫ2317/Ϫ2207; to obtain Ϫ2257/Ϫ2188 it contained Ϫ2317/ Ϫ2188, and to obtain Ϫ2257/Ϫ2172 it contained Ϫ2317/Ϫ2172. Point mutations were introduced into the 163-bp Sau3AI fragment using the Altered Sites II in vitro Mutagenesis System (Promega). In the mutant oligos that follow, substitutions in the wild-type sequence are indicated in boldface: GREm (Ϫ2253 GTGATTTCAGGCGTGGACTCTGTACTT), NF1m1, synthesized by Michel Lambert of this Research Center (Ϫ2218 TTGGCACAGTGCTTCCATCAACTTGA), NF1dm2 (Ϫ2224 CT-GACCTTGGTACAGTGCTTC), HXm (Ϫ2232 GTACTTTCCTGAGATT-GGCACAGT), HXm-NF1dm2 (Ϫ2232 GTACTTTCCTGAGATTGGTA-CAGT). Oligo HXm-NF1dm2 contains, in addition to the HXm mutation, a single mutation in the NF1 sequence; since it was used to mutate a Sau3AI fragment that already carried NF1 m1, it generated a triple mutant sequence. A 3-bp tandem point mutation was also introduced into the AAAG core of the Barbie box of the CYP2B2 promoter using oligo BBm (Ϫ94 AGTGAATAGCCAGCTCAGGAGGCGTGA). Deletion and point mutants were subcloned by blunt-ended ligation into the EcoRV site of the non-PB responsive Ev construct which contains 1681 bp of the CYP2B2 5Ј-flanking region cloned upstream of the cat reporter of pBSCAT (9). The strategy employed to obtain deletions by Bal31 digestion and by amplification generated an additional 8-nucleotide sequence (5Ј-GGGGGATC) or, for the deletions obtained with oligo HX, and additional 12-nucleotide sequence (5Ј-GATCGGGGGATC), between the 3Ј end of the deleted portion of the 163-bp Sau3AI fragment and the EcoRV site used for subcloning. The normal orientation and integrity of the subcloned fragments in the reporter constructs were confirmed by DNA sequence analysis.
Nuclear Extracts and DNase I Footprinting Analyses-Nuclear extracts were prepared (22) from pooled livers of two or three untreated or PB-treated (23) rats. Where noted, nuclear protein extracts were fractionated by chromatography on heparin-Sepharose using Hitrap Heparin columns (Amersham Pharmacia Biotech). The fragments used for DNase I footprinting analyses were prepared from pSa-Sa163, or from similar plasmids containing a mutated 163-bp Sau3AI fragment, by releasing a 230-bp fragment with EcoRI and EagI digestions. After 3Ј end-labeling using the Klenow fragment of Escherichia coli DNA polymerase I or 5Ј end-labeling using T4 polynucleotide kinase, the fragment was purified by electrophoresis on a 5% polyacrylamide gel. DNase I footprinting was performed (22) in 12.5 mM Hepes, pH 7.6, 70 mM KCl, 3.5 mM MgCl 2 , 5 M ZnSO 4 , 1 mM EDTA, 5% glycerol, 0.5 mM NaMoO 4 , 0.075 mM Nonidet P-40, 0.5 mM dithiothreitol, 0.25 mM phenylmethylsulfonyl fluoride containing 5 g of poly(dI-dC). Reaction vol-umes were typically 70 l. Na 2 ␤-glycerophosphate, when present, was at 10 -20 mM.

Functional Analysis of the 163-bp Sau3AI
Fragment-Transfection analysis was performed with subfragments of the 163-bp Sau3AI fragment to identify sequences within it that might confer PB responsiveness. Sequential deletions from the 5Ј end up to the internal NcoI site (coordinate Ϫ2257) were all active in conferring PB responsiveness, although the response conferred by construct Ϫ2257/Ϫ2155 was reduced by about 2-fold as compared with that of the full-length fragment (Fig.  1A, constructs with a common 3Ј end point of Ϫ2155 and 5Ј end points of Ϫ2290, Ϫ2283, Ϫ2273, Ϫ2264, or Ϫ2257). Further deletion from the 5Ј end, to an internal RsaI site (coordinate Ϫ2230) or to Ϫ2180, led to complete loss of activity (Fig. 1A). Similarly, sequential 3Ј deletions up to Ϫ2207 were also active, although the response conferred by constructs Ϫ2317/Ϫ2188 and Ϫ2317/Ϫ2207 was again reduced by about 2-fold as compared with that of the full-length fragment (Fig. 1A, constructs with a common 5Ј end point of Ϫ2317 and 3Ј end points of Ϫ2172, Ϫ2188, or Ϫ2207). Further deletion from the 3Ј end, to the RsaI site (coordinate Ϫ2231; Fig. 1A) or to Ϫ2253 (construct Ϫ2600/Ϫ2253; data not shown), led to complete loss of activity.
These results suggested that DNA sequence elements that are essential for PB responsiveness are localized in the 51-bp central core between Ϫ2257 and Ϫ2207. However, construct Ϫ2257/Ϫ2208 was completely inactive (Fig. 1), although its 5Ј extremity is the same as that of the active Ϫ2257/Ϫ2155 construct and its 3Ј extremity is 1 bp upstream of that of the active Ϫ2317/Ϫ2207 construct. Since construct Ϫ2257/Ϫ2207 was also inactive (Fig. 1B), the single base pair difference at the 3Ј ends of the Ϫ2257/Ϫ2208 and Ϫ2317/Ϫ2207 constructs does not account for the inactivity of the former. Hence, the central core is insufficient by itself to elicit PB responsiveness.
Functional Evidence for an Accessory Element-The deletion analysis presented above suggests the following model to account for the PB responsiveness conferred by the 163-bp Sau3AI fragment. A central core, together with an element or elements extending upstream of coordinate Ϫ2257 or with an element or elements extending downstream of coordinate Ϫ2207, can confer PB responsiveness. In support of this model, although construct Ϫ2257/Ϫ2188 was inactive, construct Ϫ2257/Ϫ2172 was active in eliciting a PB response (Fig. 1B). This result indicated that there is a DNA sequence element between coordinates Ϫ2188 and Ϫ2172 that confers PB responsiveness when combined with the central core. The element between Ϫ2188 and Ϫ2172 is insufficient to elicit PB responsiveness by itself, because it is present in the inactive Ϫ2230/ Ϫ2155 construct (Fig. 1A). Hence, it is an accessory site for conferring PB responsiveness, and we have designated it as AF1 (Figs. 1B and 2).
Physical Evidence for Proteins Binding to AF1 and to Other Sites within the 163-bp Sau3AI Fragment-To ascertain whether an accessory factor binds to the AF1 site, as well as to look for evidence of other protein-DNA interactions, the 163-bp Sau3AI fragment was subjected to DNase I footprinting analysis using rat liver nuclear extracts. First, the fragment was labeled on the lower strand at the 3Ј end (Fig. 3A). Analysis using crude extract revealed a series of almost continuous footprints over some 85 bp extending from about Ϫ2285 to about Ϫ2200 (Fig. 3A). Use of heparin-Sepharose-fractionated extracts facilitated resolution of protected regions F1, F2, F3, and F4 within this segment: protected region F2 was virtually undetectable with the fractionated extracts, whereas F1, F3 and F4 remained visible (Fig. 3A). An additional protected region, F0, was revealed after 5Ј end-labeling of the lower strand (Fig. 3B). F0 is also visible above the F1 footprint in Fig.  3A (crude extract lane). Protected regions F1Ј, F2Ј, and F3Ј, corresponding to F1, F2, and F3 plus F4, respectively, were identified by labeling the 5Ј end of the upper strand (Fig. 3C). Protected region F0Ј (corresponding to F0) is also evident above F1Ј in Fig. 3C. The positions of the F0 and F0Ј footprints overlap with the AF1 site defined by transfection analysis (Fig. 2).
Addition of PB to incubation mixtures did not appreciably change the footprints obtained with crude nuclear extracts of untreated rats (data not shown) or with a nuclear extract partially purified by heparin-Sepharose fractionation (Fig. 3A, lanes PB), and similar footprints were generated by crude nuclear extracts from untreated and PB-treated rats ( Fig. 3B and data not shown).
The 163-bp CYP2B2 Sau3AI Fragment Is a Multicomponent Enhancer-The DNA sequence of the 163-bp Sau3AI fragment (13) (see also Fig. 2) reveals that it contains putative recognition sites for several transcription factors, only some of which are shown in Fig. 2. Potential regulatory motifs were identified by inspection and by application of MatInspector (matrix similarity threshold, 0.85), a search tool for scanning DNA sequences for matches to nucleotide distribution matrices for transcription factor binding sites accessible in the TRANSFAC data base (24). Immediately adjacent to a perfectly symmetrical NF1 site (Ϫ2217 TGGN 7 CCA) (25), previously identified (9, 13), is a perfect hexamer half-site (AGGTCA Ϫ2223, lower strand) for orphan members of the nuclear receptor superfamily (26,27). We refer to the combined hexamer half-site and NF1 site (Fig. 2) as the HX⅐NF1 complex. There are also, in an unusual everted repeat arrangement with a 7-bp spacing (ER-7), two other AGGTCA sequences (coordinates Ϫ2282 to Ϫ2264) (Fig. 2). In addition, between the internal NcoI site and the HX⅐NF1 complex, MatInspector finds on each strand a match to a glucocorticoid receptor binding site (TRANSFAC matrix GR_Q6; coordinates Ϫ2244 to Ϫ2226 on the upper strand and Ϫ2246 to Ϫ2228 on the lower strand) (Fig. 2). The 51-bp central core between Ϫ2257 and Ϫ2207 defined by the transfection analyses described above excludes ER-7 but includes the candidate glucocorticoid receptor binding sites and most of the HX⅐NF1 complex.
The F1 and F1Ј footprints extend from Ϫ2230 to Ϫ2192, a region which includes and extends beyond the HX⅐NF1 complex (Fig. 2). Most of the F1 footprint was eliminated by a CTF/NF1 consensus oligo competitor (Fig. 3A, lane NF1). Therefore, it is caused in part at least by an NF1 protein. The positions of F3/F4 and F3Ј correspond to the two halves of the ER-7 sites, and both F3 and F4 were competed by oligo HX (Fig. 3A, lane  HX). Hence F3/F4 and F3Ј are presumably caused by protein(s) of the nuclear receptor superfamily binding to the two halves of ER-7. No good candidates for the protein(s) responsible for the F2 and F2Ј footprints are yet available.
The region of the F1Ј footprint was analyzed using varying amounts of nuclear protein. At lower protein levels, a reduced protected region was observed with the wild-type sequence, presumably as a result of binding of an NF1 protein, whereas at higher protein levels, the footprint was extended in both directions (Fig. 4A). Similar results were obtained with protein extracts from untreated and from PB-treated rats (Fig. 4A). When a mutant sequence in which NF1 binding was abolished was analyzed, the extended protection on the upstream side was still evident at higher protein levels (Fig. 4B). Thus, the extended protection on the upstream side is due to binding of a protein or proteins to a sequence including the nuclear receptor hexamer half-site. Consistent with this conclusion is the obser-vation that only a part of the F1 (Figs. 3A and 5A) and the F1Ј (Fig. 5B) footprints were eliminated by competition with an NF1 consensus oligo.

PB Responsiveness Is Reduced but Not Eliminated by Mutating Either or Both Elements of the HX⅐NF1 Complex-To inves-
tigate the role of the HX⅐NF1 complex in conferring PB responsiveness, the NF1 site was mutated in two steps as follows: first in the distal portion (NF1m1; Fig. 6) and then in both the distal and proximal portions (NF1dm2; Fig. 6). The NF1 m1 mutation created a new DNase I-hypersensitive site at Ϫ2207 and modified the F1 footprint but did not abolish it; furthermore, the portion of the modified F1 footprint corresponding to NF1 sequences was eliminated by competition with an NF1 consensus oligo (Fig. 5A). With the NF1dm2 mutant sequence, the F1Ј footprint was reduced to that seen with the wild-type sequence in the presence of an NF1 consensus oligo competitor (Fig. 5B). Hence, the NF1dm2 mutation completely eliminated detectable NF1 binding, but it left a footprint on the upstream side, corresponding to the anticipated binding of a protein or proteins to a sequence including the nuclear receptor hexamer half-site (Figs. 4B and 5B). Here, as elsewhere, identical results were obtained using nuclear extracts prepared from livers of untreated and PB-treated rats (Fig. 5B).
The NF1 m1 and NF1dm2 mutations, as well as two additional mutations of the HX⅐NF1 complex, HXm (mutated in the hexamer half-site) and HXm-NF1dm2 (mutated in the hexamer half-site and in NF1), all reduced but did not abolish PB responsiveness when tested by transfection analysis (Fig. 6). None of these mutations increased the basal level of CAT activity (data not shown). These results demonstrated that both elements of the HX⅐NF1 complex must be intact to elicit a maximal response to PB.  (13), except that our cloned sequence has three not four T residues immediately 5Ј of nucleotide Ϫ2225 (vertical arrowhead). The sequence of the CYP2B2 fragment is identical to that of the corresponding CYP2B1 fragment (44) except for a single base difference at CYP2B2 nucleotide Ϫ2299 (asterisk), which is C in CYP2B2 and G in CYP2B1. Protected regions on the upper and lower strands are shown by upper and lower brackets, respectively. The Sau3AI restriction sites and the internal NcoI site are boxed. Nuclear receptor hexamer half-sites in the ER-7 arrangement and forming part of the HX⅐NF1 complex are identified by horizontal arrows; the NF1 site is overlined, and the sequence of the AF1 site is shaded. Candidate glucocorticoid receptor binding sequences on the upper and lower strands are denoted by a dotted overline and a dotted underline, respectively. The central core sequence between Ϫ2257 and Ϫ2207 is delimited by the region between the two vertical lines. strand, matches the consensus at 7 of 12 positions. This sequence was mutated to GGCGTGgacTCTGTA, thereby reducing the match to 4 of 12. This led to virtual abolition of the PB response (Fig. 6). This suggests that a protein binding in the region of the putative GRE is required to confer PB responsiveness.

Mutation of a Candidate Glucocorticoid Receptor Binding
Mutation of the CYP2B2 Barbie Box Does Not Affect PB Inducibility-It has been suggested that the Barbie box (29) is involved in conferring PB responsiveness on the rat CYP2B1 and CYP2B2 genes (30). However, when the core sequence (5Ј-AAAG-3Ј) of the CYP2B2 Barbie box was mutated, PB responsiveness was retained (Fig. 6). DISCUSSION Our previous work has shown that an upstream enhancer element, located between Ϫ2317 and Ϫ2155 in 5Ј-flanking region, confers PB responsiveness on the rat CYP2B2 gene (9). This element is situated in the vicinity of a liver-specific DNase I-hypersensitive site in chromatin (10). Such sites are a hallmark of regulatory regions (31). Furthermore, in transgenic mice a rat CYP2B2 transgene including only the first 800 bp of the 5Ј-flank was not PB-inducible, whereas a transgene carrying 19 kb of 5Ј-flank (and hence the Ϫ2317/Ϫ2155 segment) was normally inducible (32). The experiments described in our previous report (9) and in subsequent reports from two other laboratories (11,12), as well as those described here, provide a strong body of evidence indicating that upstream control elements confer PB responsiveness on the rat CYP2B2 and mouse Cyp2b10 genes. The mouse Cyp2b10 active sequences are situated between Ϫ2426 and Ϫ2250 and display 91% sequence identity to those of CYP2B2 (12). Particularly important in this context is the observation that generally similar results have been obtained in three different laboratories using different experimental approaches and PB-inducible CYP2B genes of two different rodent species.
The results presented here, showing that at least three and probably more sequence elements within the 163-bp Sau3AI fragment are required to confer maximal PB responsiveness, FIG. 3. Binding of liver nuclear proteins to the CYP2B2 PBRU. DNase I footprinting assays were performed as described under "Experimental Procedures." Protected regions F0 through F4 on the lower strand and F1Ј through F3Ј on the upper strand are represented by boxes. BSA denotes DNase I treatment in the absence of nuclear proteins. A, the fragment was 3Ј end-labeled on the lower strand and incubated with a heparin-Sepharose fraction eluted at 0.7 M KCl (50 g of protein) or with 100 g of unfractionated nuclear protein (denoted crude extract) from untreated rats. ϩNF1 and ϩHX denote the presence of a 500-fold molar excess of the CTF/NF1 consensus oligo or of the double-stranded HX oligo competitors, respectively, and ϩPB denotes preincubation of fractionated proteins with 4 mM PB (left lane, 1.5 h on ice without the probe and right lane, 2 h on ice with the probe). Protected regions corresponding to proteins binding to the NF1 and ER-7 sites and to the hexamer half-site, as determined from Maxam-Gilbert sequencing ladders (not shown), are indicated by brackets labeled NF1, ER7, and HX, respectively. B, the fragment was 5Ј end-labeled on the lower strand and was treated with unfractionated nuclear extract (100 g of protein). GϩA denotes the Maxam-Gilbert sequencing ladder, and control and PB denote DNase I treatment with nuclear extracts from untreated and PB-treated rats, respectively. C, the fragment was 5Ј end-labeled on the upper strand and was treated with unfractionated nuclear extract (100 g of protein), denoted control. reveal the surprising complexity of this multicomponent enhancer. According to the now classical model for steroid hormone action, the liganded receptor binds to one or more response elements upstream of the target gene and acts as a transcriptional enhancer (26). However, synergistic interaction of a variety of transcription factors with steroid hormone response elements have been known for some time (33), and in a number of genes several cis-acting elements function in concert to activate transcription (34). The best studied such case is that of the rat PEPCK gene. Full glucocorticoid induction of PEPCK transcription is mediated by a complex GRU consisting of two glucocorticoid response elements, as well as three accessory factor binding sites (14,17). We originally referred to the CYP2B2 Sau3AI fragment as a PB-responsive element (9), but, in the light of its complexity, we now refer to it as a PBRU (35,49). The homologous mouse Cyp2b10 sequence has been termed a module (12,36). Proteins that can serve as accessory factors in the PEPCK glucocorticoid response include hepatocyte nuclear factor-3 (16) and the orphan receptors chicken ovalbumin upstream promoter transcription factor (15,17) and hepatocyte nuclear factor-4 (HNF-4) (15). The application of the PEPCK GRU model to PB regulation of CYP2B2 does not imply that any particular recognition site is shared between the two response units, but merely that PB induction of CYP2B2 requires interactions among multiple regulatory proteins and cis-acting elements constituting a PBRU. DNase I footprinting assays were performed using unfractionated rat liver nuclear extract. The fragment was 5Ј end-labeled on the upper strand. Only the region of the F1Ј footprint is shown. BSA denotes DNase I treatment in the absence of nuclear proteins, and control and PB denote DNase I treatment with nuclear extracts from untreated and PB-treated rats, respectively. Brackets labeled NF1 and HX denote protected regions caused by proteins binding to the NF1 site and to a sequence including the nuclear receptor hexamer half-site, respectively. A, the wild-type fragment was used. Incubations were performed in the presence of 20 mM sodium ␤-glycerophosphate, and the amounts of nuclear extract protein per assay were 2.5, 5, 7.5, 10, 25, 50, 75, and 100 g. B, the fragment carrying the NF1dm2 mutation was used. Incubations were performed in the presence of 10 mM sodium ␤-glycerophosphate, and the amounts of nuclear extract protein per assay were 2.5, 5, 7.5, 10, 25, 50, and 75 g.

FIG. 5. Effect of mutations in the NF1 site on the F1 and F1 footprints.
A, the wild-type fragment or the fragment carrying the NF1 m1 mutation was treated with unfractionated nuclear extract (100 g of protein) from livers of untreated rats. Fragments were 3Ј end-labeled on the lower strand. NF1 denotes the presence of a 500-fold molar excess of the NF1 consensus oligo competitor. The position of the F1 footprint is shown by a box, and HS at coordinate Ϫ2207 denotes a hypersensitive site. B, the wild-type fragment or the fragment carrying the NF1dm2 mutation was used. Fragments were 5Ј end-labeled on the upper strand. BSA denotes DNase I treatment in the absence of nuclear proteins, and control and PB denote DNase I treatment with nuclear extracts (100 g of protein) from untreated and PB-treated rats, respectively. NF1 denotes the presence of a 500-fold molar excess of the NF1 consensus oligo competitor. The position of the F1Ј footprint is shown by a box and that of the footprint remaining after NF1 binding was eliminated by competition or mutation is denoted by a bracket labeled HX.
The essential element of the PBRU model then is that more than one sequence element within the 163-bp Sau3AI fragment is required to confer PB responsiveness. From our results we conclude that a central core of the fragment, between coordinates Ϫ2257 and Ϫ2207, is inactive alone but can confer PB responsiveness when combined either with upstream or downstream accessory elements within the fragment. In support of this hypothesis, one such accessory site, designated AF1, has been localized between coordinates Ϫ2188 and Ϫ2172. AF1 presumably represents an accessory factor binding site. The protein(s) responsible for the F0/F0Ј footprint may also be responsible for AF1 activity, since these footprints overlap with the AF1 sequence.
The level of PB responsiveness of construct Ϫ2257/Ϫ2172 (Fig. 1B) is essentially the same as that of construct Ϫ2257/ Ϫ2155 (Fig. 1A), and both are about 2-fold lower than the wild-type Ϫ2317/Ϫ2155 construct. This indicates that sequences between Ϫ2172 and Ϫ2155 are not essential for maximal PB responsiveness and, moreover, suggests that a second accessory element may be present between Ϫ2317 and Ϫ2257. Since full PB responsiveness is retained in 5Ј deletion constructs up to Ϫ2264, the putative second accessory element may lie between Ϫ2264 and Ϫ2257. However, it is also possible that the reduced responsiveness of the Ϫ2257/Ϫ2155 construct (and of the Ϫ2257/Ϫ2172 construct) is a consequence of partial inactivation of essential sequences at the 5Ј end of the central core. Experiments are currently underway to resolve these issues. In any case, removing one (or the other) of the putative CYP2B2 PBRU accessory sites reduces but does not abolish PB responsiveness. Similarly, eliminating any one of the three PEPCK GRU accessory sites reduces but does not abolish GRUmediated glucocorticoid responsiveness (17).
Our results indicate that the HX⅐NF1 complex is required for maximal PB responsiveness (Fig. 6). NF1 proteins are abundant, ubiquitous transcription factors (37), different isoforms of which are products of a multigene family (38). The NF1 consensus binding sites are composed of two motifs, TGG and GCCAA, separated by a 6-or 7-bp spacer, and the protein protects a 25-30-bp region surrounding this sequence from DNase I digestion (25). An NF1 protein is clearly responsible for part of the F1/F1Ј footprint and the role of NF1 is positive, because mutations which reduce (e.g. NF1 m1) or abolish (e.g. NF1dm2) NF1 binding reduce but do not abolish PB responsiveness (Fig. 6). This result is similar to that of Honkakoski and Negishi (12) with a different NF1 mutation that reduced but did not eliminate PB responsiveness conferred by a subfragment of the mouse Cyp2b10 homolog of the CYP2B2 163-bp Sau3AI fragment. In our hands, mutations of the nuclear re- Transfection experiments as well as analysis and presentation of CAT assay results were as described in the legend to Fig. 1. DNA sequences within the shaded panel at left represent the wild-type F1Ј-protected region, denoted WT, and various mutants of the HX⅐NF1 complex as shown. The mutated versions of the putative GRE and of the Barbie box are also shown at left as well as their sequence coordinates. For all mutant sequences, the upper strand is shown, and substitutions in the wild-type sequence are underlined. The F1Ј-protected sequence is shown in double-stranded form at the bottom. Brackets labeled HX and NF1 delimit the footprints attributed to proteins binding to the hexamer half-site and to the NF1 site, respectively. ceptor half-site element, as in HXm, or of both that element and NF1, as in HXm-NF1dm2, also reduce PB responsiveness, that is they have similar effects to those of mutational inactivation of NF1 alone (Fig. 6). In the mouse Cyp2b10 system, Honkakoski and Negishi (12) reported a more dramatic effect of such mutations in reducing PB responsiveness. Hence, the HX⅐NF1 complex is clearly essential for conferring maximal PB responsiveness.
The protein binding to the hexamer half-site creating the HX portion of the F1/F1Ј footprint is unknown, but there are several possibilities. Inspection of the sequence surrounding the HX⅐NF1 complex reveals a recognition site for the orphan nuclear receptor, fetoprotein transcription factor (22,39), and MatInspector finds a match to recognition sites for three other orphan receptors, including chicken ovalbumin upstream promoter transcription factor and HNF-4. Preliminary experiments have revealed binding of both HNF-4 (40) and fetoprotein transcription factor 2 to the double-stranded HX oligo. Which of these or other proteins interacts functionally with the HX⅐NF1 complex to elicit maximal PB responsiveness remains to be determined.
The nature of the proteins responsible for the F0/F0Ј and F2/F2Ј footprints is currently unknown. They are of particular interest because F2/F2Ј overlaps the central core identified as essential for PB responsiveness whereas F0/F0Ј overlaps the accessory AF1 site.
The protein responsible for the F3-F4/F3Ј footprints is presumably a member of the nuclear receptor superfamily, because of the overlap with ER-7 site and because the F3-F4 footprint is competed by the HX oligo containing the AGGTCA hexamer half-site. The ER-7 region is not required for PB responsiveness, so it is clearly not an essential site, although it may be an accessory site. Similar conclusions were reached by Honkakoski and Negishi (12) with regard to the corresponding Cyp2b10 pB sequence.
Our observation that mutation of a putative GRE leads to a major reduction in PB responsiveness is particularly intriguing, as it raises the possibility that a glucocorticoid receptorlike molecule may be involved in conferring PB responsiveness. The putative PBRU GRE matches the GRE consensus at only 7 of 12 positions. Hence, like the GREs forming part of the PEPCK GRU (14), it is not a typical GRE.
Much previous effort has been concentrated on promoterproximal sequences, particularly the Barbie box (29,30) and their possible role in conferring PB responsiveness on the rat CYP2B1 and CYP2B2 genes. However, mutation of the Barbie box sequence in the context of the CYP2B2 promoter and 5Јflank does not affect PB responsiveness either in rat hepatocytes transfected by injection in situ (11) or in transfected primary rat hepatocytes (Fig. 6). Hence, while promoter-proximal sequences are doubtless required for basal transcriptional activity (9,36), there is at present no convincing evidence for their involvement in conferring PB responsiveness.
The nucleotide sequences of the coding regions of CYP2B1 and CYP2B2 are about 97% identical (41,42), and this extends over at least 1 kb of their 3Ј-flanking regions (43) and over some 2.3 kb of their 5Ј-flanking regions (13,44). This, plus the fact that the responses of liver CYP2B1 and CYP2B2 to PB treatment are very similar (although the basal level of CYP2B2 is somewhat higher than that of CYP2B1 (45,46)), explains why CYP2B1 and CYP2B2 are sometimes treated as though they were a single gene (47,48). Nevertheless, there are striking differences in the tissue-specific expression of CYP2B1 and CYP2B2, most notably in lung where CYP2B2 is not expressed, whereas CYP2B1 is expressed constitutively but is not PBinducible (46). Because the 163-bp Sau3AI fragment is similarly positioned and, except for a single nucleotide difference, identical in sequence in CYP2B1 and CYP2B2 (13, 44) (see also Fig. 2), we will have to look elsewhere to explain the differences between CYP2B1 and CYP2B2 expression in the lung. It seems likely that the CYP2B2 PBRU and CYP2B1 PBRU are both inactive in the lung and that CYP2B1 expression is turned on because of the constitutive presence of some other transcription factor, the activity of which depends on one of the subtle sequence differences between the CYP2B1 and CYP2B2 5Ј-flanking regions.
The molecular details as to how the presence of PB leads to activation of transcription via the PBRU remain to be determined. However, a recent report of results obtained by in vivo footprinting suggests that PB may modify the interaction of transcription factors with the PBRU in chromatin (48). Although this provides further support for the role of the PBRU in transcriptional activation, further characterization of the proteins that bind to it will be necessary to elucidate the mechanism.

Phenobarbital Response Unit Contains an Accessory Factor Element CYP2B2
The