The Human Prepro-orexin Gene Regulatory Region That Activates Gene Expression in the Lateral Region and Represses It in the Medial Regions of the Hypothalamus*

Prepro-orexin is a precursor of the neuropeptides orexin-A and -B, which are localized in the neuronal population of the lateral hypothalamic area (LHA). We wished to elucidate the mechanisms by which the prepro-orexin gene is specifically activated in orexin neurons in the LHA. The 3.2-kb 5′-flanking region of the human prepro-orexin gene is sufficient for the specific expression of anEscherichia coli lacZ reporter gene in orexin neurons. Therefore, we examined a series of reporter constructs harboring this 3.2-kb regulatory region or its deletion in a reporter transgenic mouse assay. There are two phylogenetically conserved regions located 287 bp (orexin regulatoryelement (OE) 1) and 2.5 kb (OE2) upstream of the transcription initiation site of the human prepro-orexin gene. In transgenic mice, both OE1 and OE2 are necessary for expressing the human prepro-orexin gene in the LHA and for repressing its expression in the medial regions of the hypothalamus. Through serial deletion analysis of OE1, we found that the 57-bp core region of OE1 is critical for its spatial gene regulatory function in vivo. Mutation analysis further demonstrated that without contribution from the OE1 core region, the lacZ reporter is expressed ectopically in the medial regions of the hypothalamus. Thus, OE1 contains crucialcis-acting elements regulating prepro-orexin gene expression specifically in the LHA.

modulating body weight and metabolism (3,4). Several lateral hypothalamic neuropeptides controlling the regulation of energy balance and food intake have also been identified (5). For instance, orexin-A (hypocretin-1) and orexin-B (hypocretin-2) and their common precursor peptide prepro-orexin are specifically localized in neurons located in the lateral hypothalamic area (LHA), a region classically regarded as the "feeding center" (6 -8). These neuropeptides are ligands for two closely related orphan G-protein-coupled receptors, OX1r and OX2r (8,9). OX1r is selective for orexin-A, whereas nonspecific OX2r binds both isopeptides. In rats, intracerebroventricular administration of orexins stimulates food consumption, whereas the expression level of mRNA is up-regulated by fasting (8) and decreased in genetically obese ob and db mice (10). The expression level of the prepro-orexin gene is regulated by leptin and orexin-containing neurons express the leptin receptor (11).
It is believed that orexins also play roles in regulating states of arousal, sleep, and wakefulness. For example, it was reported that targeted disruption of the mouse prepro-orexin gene results in a sleeping disorder strikingly similar to human narcolepsy (12). Furthermore, mutations of the OX2 gene were found in a canine narcolepsy model (13). Prepro-orexin gene expression is highly restricted to a specific population of neurons located in the LHA. We wanted to expound the gene regulatory mechanism through which the prepro-orexin gene specifically activates transcription in such neurons. This information should yield further insights into the physiological function of orexins, the precise mode of gene regulation of various hypothalamic neuropeptides, and the mechanism of development of the hypothalamus.
In transgenic mice, the 3.2-kb upstream region of the human prepro-orexin gene is sufficient to direct the expression of an Escherichia coli lacZ reporter gene in orexin-immunoreactive neurons in the LHA (14). Furthermore, expressing a toxic transgene using the same 3.2-kb fragment abolishes orexin neurons without affecting any other neurons (15). Thus, this 3.2-kb genomic fragment contains all the critical cis-acting elements required for prepro-orexin gene expression in the LHA. To further reveal the gene regulatory regions, we determined the evolutionarily conserved common nucleotide sequences in the 5Ј-flanking regions of the human (14) and mouse (12) prepro-orexin genes. This "phylogenetic footprinting" approach (16,17) revealed that there are two phylogenetically conserved regions located 287 bp (orexin regulatory element (OE) 1) and 2.5 kb (OE2) upstream of the transcription initiation site of the human prepro-orexin gene. In transgenic mice, both OE1 and OE2 are necessary for expressing the human prepro-orexin gene in the LHA and for repressing its expres-sion in the medial regions of the hypothalamus. Through serial deletion analysis of OE1, we delineated the 57-bp core region within the human OE1 sequence. This core region was indispensable for lacZ reporter gene expression in orexin-immunoreactive (orexin-ir ϩ ) neurons in the LHA of transgenic mice. These results thus demonstrate that OE1 is a crucial region regulating prepro-orexin gene expression specifically in the LHA.

EXPERIMENTAL PROCEDURES
Computer Analysis of DNA Sequences-A 5.1-kb KpnI fragment of the 5Ј-flanking region of the mouse prepro-orexin gene was prepared. This 5.1-kb sequence was subcloned into the pBluescript SK-II(ϩ) vector and sequenced using a PerkinElmer Life Sciences Model 377 DNA sequencer. A multiple nucleotide sequence alignment was performed between the 5Ј-flanking regions of the mouse (this study) and human (14) prepro-orexin genes using GENETYX-MAC Version 9.0 (Software Development Co., Tokyo). A search for transcription factor recognition sequences was performed with TFSEARCH. 2 Construction of Plasmids and Generation of Transgenic Mice-We made reporter constructs containing a lacZ gene and an SV40 T-antigen nuclear localization signal ligated to either the entire or truncated regions of the 3.2-kb regulatory domain of the human prepro-orexin gene (14). All deletion mutants were generated using a multistep PCRbased approach. 3 Both strands of the plasmids were sequenced to confirm the authenticity of the constructs. Transgenic mice were generated by standard methods (18). Because the level of orexin expression gradually increases during postnatal development, analysis was performed using transgenic founder mice at 4 -9 weeks of age (19).
Quantitative Analysis of ␤-Galactosidase-positive Neurons within Orexin-ir ϩ Neurons in the LHA-We examined lacZ reporter gene expression in the hypothalamuses of transgenic founder mice (F 0 or G 0 for generation 0) from 4 to 9 weeks of age throughout this study. The percentage of the ␤-galactosidase-positive (␤-gal ϩ ) population within the orexin-ir ϩ neurons in the LHA was determined by counting ␤-gal ϩ neurons and orexin-ir ϩ neurons in 10 serial coronal sections containing ϳ2 ϫ 10 3 orexin neurons. The percentage of ␤-gal ϩ neurons distributed in the medial regions of the hypothalamus was calculated relative to the number of the total ␤-gal ϩ neurons in the whole hypothalamus in the same coronal sections. The percentage was used as a parameter for ectopic medial expression of the reporter gene.
Electrophoretic Mobility Shift Assay-Electrophoretic mobility shift assay (EMSA) was performed as described previously (20,21) using four oligonucleotide probes covering the core region of OE1 (Ϫ258 to Ϫ201 bp). Nuclear extract was prepared from mouse hypothalamus following a standard procedure (22). Hypothalamic nuclear extracts (10 g) were incubated at room temperature for 30 min in a volume of 25 l containing 100 pg of 32 P-labeled double-stranded oligonucleotide probe, 2 g of poly(dI-dC)⅐poly(dA-dT), 75 mM KCl, 10 mM Tris-HCl, 1 mM EDTA, 1 mM dithiothreitol, and 4% Ficoll. DNA-protein complexes were separated by 4% nondenaturing polyacrylamide gel electrophoresis and analyzed by autoradiography.

Human and Mouse Prepro-orexin Genes Contain Two Conserved Regulatory
Regions-Because we wished to study the basic mechanisms of prepro-orexin gene transcriptional regulation, we searched for regions conserved during molecular evolution in orexin genes. To this end, we determined the nucleotide sequence of the 5.1-kb KpnI fragment containing the 5Ј-flanking region of the mouse prepro-orexin gene. Sequence comparisons were then made of the 5Ј-flanking regions of the transcription initiation sites of the mouse (this study) and human (14) prepro-orexin genes, respectively. Two highly conserved regions were identified and called OE1 and OE2 (Fig. 1,  A and B). OE1 encompasses a 214-bp PvuII-PstI region located 287 bp 5Ј of the transcription initiation site of the human prepro-orexin gene. OE2 is a 217-bp region and lies 2.5 kb upstream of the transcription start site of the human gene. No conserved regions were found in either the introns or the 3Јflanking regions of the human and mouse prepro-orexin genes.
FIG. 1. Human prepro-orexin gene regulatory regions. A, two highly conserved regions, termed OE1 (gray ovals) and OE2 (black ovals), exist in the upstream regions of the human and mouse prepro-orexin genes. Hatched circles indicate Alu repeats. B, shown are the nucleotide sequences of the human and mouse OE1 elements. Asterisks indicate nucleotides that are conserved between the human and mouse prepro-orexin genes. Nucleotide residues are numbered negatively, starting at the transcription initiation site of the human prepro-orexin gene.
In the human gene, five copies of the highly repetitive elements of the primate Alu family are located in the middle of the 3.2-kb upstream region (14). Although some mouse B1 repeats, which are counterparts of the human Alu repeats, were observed in the mouse 5Ј-flanking region, these repetitive sequences were not involved in the phylogenetically conserved regions (23).

OE1 Is Necessary for Specific Gene Expression in Orexin
Neurons in the LHA-The 3.2-kb upstream promoter region of the human prepro-orexin gene contains OE1, OE2, and five copies of an Alu repeat sequence. To delineate which regulatory modules direct LHA-specific orexin gene expression, we examined a series of constructs containing deleted regions of the 3.2-kb sequence fused to a lacZ reporter gene (Fig. 2). Consistent with the previous analysis (14), the complete 3.2-kb regu-latory region (3.2nlacZ) drove lacZ gene expression specifically in orexin-ir ϩ neurons in the LHA (Fig. 3A). Indeed, five of eight 3.2nlacZ transgenic founder mice showed ␤-gal ϩ staining ( Fig.  2) that overlapped with anti-orexin immunostaining (Fig. 3B).
To identify the importance of the gene regulatory activity of OE1, we generated a construct lacking the OE1 sequence (delt-OE1nlacZ) by making an internal deletion (Ϫ287 to Ϫ77 bp) mutation of 3.2nlacZ (Fig. 2). Surprisingly, no ␤-gal ϩ staining was observed in orexin-ir ϩ neurons residing in the LHA of delt-OE1nlacZ transgenic mice. In contrast, three of eight delt-OE1nlacZ transgenic founder mice showed ␤-gal ϩ staining in the medial regions of the hypothalamus, including the dorsomedial hypothalamus, the ventromedial hypothalamus, and the ARC (Figs. 2 and 3, C and D). These results indicate that a Orexin-ir ϩ neurons are stained brown. A shows macroscopic observation of the hypothalamus of a 3.2nlacZ transgenic mouse. Note that a symmetrical distribution of ␤-gal ϩ neurons was observed in the LHA. B shows a higher magnification of the hypothalamus of a 3.2nlacZ transgenic mouse. All ␤-gal ϩ neurons contained immunoreactive orexins (arrows). C and D show that in delt-OE1nlacZ mice, ␤-gal ϩ signals were observed in non-orexin neurons in the medial regions of the hypothalamus (dashed rectangle), including the dorsomedial hypothalamus (DMH), ventromedial hypothalamus (VMH), and ARC. E and F show that in 1.3nlacZ mice, a smaller population of ␤-gal ϩ signals was observed in orexin-ir ϩ neurons in the LHA. Ectopic ␤-gal expression was observed in the ARC. G and H show that in delt-AlunlacZ mice, highly specific ␤-gal expression was observed in orexin-ir ϩ neurons (arrows). shift in lacZ reporter gene expression from the lateral to the medial regions of the hypothalamus occurs when OE1 is deleted from the 3.2-kb regulatory region.
The minimal construct 0.4nlacZ, containing OE1 and the endogenous promoter region, failed to direct lacZ expression in orexin neurons (Fig. 2). In fact, the only lacZ gene expression observed with this transgenic construct was ectopically in the posterior hypothalamus of one (of five) transgenic mouse. Addition of a 0.9-kb upstream region (1.3nlacZ) recovered the lacZ reporter gene expression in orexin neurons in the LHA of all seven transgenic founder mice examined (Figs. 2 and 3, E and F). However, 1.3nlacZ also drove the lacZ gene expression in the ARC. A similar expression profile was observed in the 2.1nlacZ transgenic founder mice, albeit with lower frequency, indicating that the required regulatory regions for transgene expression in both the LHA and ARC are located in the 0.9-kb region. It is intriguing that most of the 0.9-kb region is composed of human Alu repeats (see Fig. 2

and below).
These observations led us to quantify the ␤-gal ϩ population within the lateral hypothalamic orexin-ir ϩ neurons (␤-gal ϩ / orexin-ir ϩ neurons) and the medial regions of the hypothalamus (Table I). In 3.2nlacZ transgenic mice, the ␤-gal reporter was expressed in 34.3% (n ϭ 5) of the orexin-ir ϩ neurons (Table  I). However, the number of these ␤-gal ϩ /orexin-ir ϩ neurons declined to 11.2% in 1.3nlacZ transgenic mice (n ϭ 7). Furthermore, ectopic lacZ gene expression in non-orexin neurons within the ARC was seen in all seven 1.3nlacZ transgenic founder mice. Calculation of the number of ␤-gal ϩ neurons in the medial regions of the hypothalamus as a percentage of the total ␤-gal ϩ cells in the whole hypothalamus showed that all ␤-gal ϩ neurons occurred in the medial regions of the hypothalamuses of delt-OE1nlacZ transgenic mice (Table I). This is in stark contrast to the insignificant lacZ reporter gene expression found in the medial regions of the hypothalamuses of transgenic mice expressing the intact 3.2nlacZ gene. In 1.3nlacZ and 2.1nlacZ mice, 22.6% (n ϭ 7) and 56.3% (n ϭ 2), respectively, of the total ␤-gal ϩ neurons were found in the medial regions of the hypothalamus (Table I), indicating that the 2.1-kb region is not sufficient to define the specific LHA expression profile associated with normal prepro-orexin gene expression.
OE2 Is Indispensable for the LHA-specific Expression of the Prepro-orexin Gene-We then examined the function of OE2 to direct the LHA-specific expression of the prepro-orexin gene with the reporter transgenic assay. To our surprise, attachment of the OE2 region to the basic 0.4nlacZ construct restored the lacZ expression specificity in the orexin neurons in the LHA of two of nine delt-AlunlacZ transgenic founder mice (Fig.  2). The expression driven by the construct lacking Alu repeats was highly specific to the orexin-ir ϩ neurons (Fig. 3, G and H). Importantly, the population of ␤-gal ϩ /orexin-ir ϩ neurons was comparable to that of the 3.2nlacZ transgenic mice, with undetectable ␤-gal staining in the medial regions of the hypothalamus (Table I). These results indicate that, in addition to OE1, OE2 is indispensable for the cell lineage-specific expression of the prepro-orexin gene in vivo.
Variation of Transgene Expression in Progeny Animals within One Line-Because our analyses so far utilized the transgenic F 0 mice exclusively, concerns developed as to whether transgene expression is stably transmitted to the next generation and whether there is a proportional level of transgene expression among progeny animals from a single line. To address these issues, some of the transgenic F 0 mice generated in this study were mated to obtain their progeny mice. As a result, 15 founder mice resulted in progeny. Of the 15 lines of progeny mice, transgene expression was observed in nine lines: 3.2nlacZ (lines A3 and D5), 1.3nlacZ (lines 459 and 460), 2.1nlacZ (lines 476 and 667), delt-AlunlacZ (line 1000), 3.2delt-1nlacZ (line 124), and 3.2delt-2nlacZ (line 20); and the expression level in the F 1 mice was comparable to that in the F 0 mice, indicating that the regulatory activity within the transgenes was transmitted stably through generations. Indeed, some of the lines were mated further up to F 9 , but ␤-gal ϩ neurons did not decrease with the generation number in these lines (data not shown).
We then compared lacZ reporter gene expression in the orexin neurons of five progeny animals per transgenic line. No significant variation of lacZ expression was observed in the five progeny animals from each line (Table II), suggesting that once established as a line, transgene expression is stably maintained. In contrast, ␤-gal ϩ neurons were not found in the other TABLE I Transgenic mouse reporter gene expression analysis of a series of deletion constructs of the 3.2-kb upstream promoter region of the human prepro-orexin gene Transgenic founder mice were used in this analysis. The percentage of the ␤-gal ϩ population within the orexin-ir ϩ neurons in the LHA was determined by counting ␤-gal ϩ neurons and orexin-ir ϩ neurons in 10 serial coronal sections containing ϳ2 ϫ 10 3 orexin neurons. The percentage of ␤-gal ϩ neurons distributed in the medial regions of the hypothalamus (HT) was calculated relative to the number of total ␤-gal ϩ neurons in the whole hypothalamus in the same coronal sections.

Construct
Transgenic mouse name ␤-gal ϩ /orexin-ir ϩ ␤-gal ϩ in medial HT/ total ␤-gal ϩ in HT  Fifteen transgenic F 0 mice resulted in transgene-positive progeny. The expression of the transgene was observed in nine lines, whereas the other six did not transmit transgenic reporter gene expression to the F 1 generation. We then examined lacZ reporter gene expression in five progeny animals per transgenic line and compared the expression profiles. The percentage of the ␤-gal ϩ population in orexin-ir ϩ neurons in the LHA was determined as described in the legend to Table I  The Core Sequence within OE1 Is Essential for Prepro-orexin Gene Expression-To identify the core region of OE1 that specifies prepro-orexin gene expression in the LHA, we prepared several deletion constructs of OE1 (Ϫ287 to Ϫ77 bp) and examined them in a series of transgenic founder mouse assays (Fig. 4A). Upon deletion of the most 5Ј 37 bp (3.2delt-1nlacZ), lacZ gene expression was observed in orexin neurons in six of seven transgenic founder mice (Fig. 4A). However, the ␤-gal ϩ / orexin-ir ϩ neuron population was rather small (9.1%, n ϭ 5), and 34.3% of the total hypothalamic ␤-gal ϩ neurons were located ectopically in the medial regions of the hypothalamus, especially in the ARC (Fig. 4B and Table III). Further deletion of the 5Ј-OE1 sequence (Fig. 4A) caused all ␤-gal ϩ neurons to locate to the ARC rather than to the LHA of 3.2delt-2nlacZ (Fig. 4C) and 3.2delt-3nlacZ (Fig. 4D) transgenic mice. This specific ARC localization was observed in most transgenic mouse lines harboring 3.2delt-2nlacZ and 3.2delt-3nlacZ (Fig.  4A). Thus, the essential gene regulatory core region of OE1 is located between 250 and 202 bp upstream of the transcription initiation site. B-E, shown are the ␤-gal expression sites in the hypothalamuses of mice containing the 5Ј-deletion OE1 constructs. The 3.2delt-1nlacZ mutant mouse showed a relatively smaller population of ␤-gal ϩ cells in orexin-ir ϩ neurons and ectopic ␤-gal expression in the ARC (B). In mice with the 3.2delt-2nlacZ (C), 3.2delt-3nlacZ (D), and 3.2deltOE1-IIn-lacZ (E) mutations, all ␤-gal ϩ signals were observed ectopically in the ARC, whereas no orexin-ir ϩ neuron-specific ␤-gal expression was observed.

TABLE III
Transgenic mouse reporter gene expression analysis of a series of OE1 deletion mutants Transgenic founder mice were used for this analysis. Percentages of the ␤-gal ϩ population in orexin-ir ϩ neurons in the LHA and of ␤-gal ϩ neurons distributed in the medial regions of the hypothalamus (HT) were determined as described in the legend to Table I.

Construct
Transgenic mouse name ␤-gal ϩ orexin-ir ϩ ␤-gal ϩ in medial HT/ total ␤-gal ϩ in HT To verify this hypothesis, we generated transgenic mice possessing an internal deletion construct lacking region Ϫ258 to Ϫ201 of OE1 (3.2deltOE1-IInlacZ) and analyzed the lacZ reporter expression. As expected, the lacZ reporter gene failed to express in the orexin-ir ϩ neurons, but expressed ectopically in the ARC (Fig. 4E and Table III). These results indicate that sequence Ϫ258 to Ϫ201 of OE1 represents the core regulatory region governing the function of OE1. We therefore designated this 57-bp sequence as the OE1 core region.
Loss of Any Quarter of the OE1 Core Affects Prepro-orexin Gene Expression-We performed a series of deletion analysis within the OE1 core region (Fig. 5A). Regions Ϫ258 to Ϫ245 (II-1), Ϫ245 to Ϫ229 (II-2), Ϫ229 to Ϫ219 (II-3), and Ϫ219 to Ϫ201 (II-4) were deleted from the OE1 core region to generate the deletion mutant reporter constructs 3.2deltII-1nlacZ, 3.2deltII-2nlacZ, 3.2deltII-3nlacZ, and 3.2deltII-4nlacZ, respectively. All transgenic founder mice carrying these deletion reporter genes showed a similar marked reduction in the population of ␤-gal ϩ /orexin-ir ϩ neurons (Fig. 5, B-F; and Table  IV), with no substantial difference being observed among the four founder mutant mice (Table IV). Such results suggest that the 57-bp OE1 core region is modular in structure, with several cis-acting modules acting cooperatively. Therefore, disruption of any single module might be compensated for by the function of the other cis-acting elements. FIG. 5. lacZ reporter gene expression sites observed in the OE1 core region analysis. A, four reporter constructs (3.2deltII-1nlacZ, 3.2deltII-2nlacZ, 3.2deltII-3nlacZ, and 3.2deltII-4nlacZ) were examined in transgenic founder mice. B, the number of founder mice showing ␤-gal ϩ staining (␤-Galϩ) and the total number of transgenic founder mice (Tgϩ) are shown. C-F, transgenic mice harboring the 3.2deltII-1nlacZ, 3.2deltII-2nlacZ, 3.2deltII-3nlacZ, and 3.2deltII-4nlacZ reporter genes, respectively, showed a marked reduction in the ␤-gal ϩ population in orexin neurons, as indicated by arrows.
FIG. 6. Transgenic mouse analysis of the substituted mutation construct 3.2OE1-IImnlacZ. A, the nucleotide sequences of the OE1 core region and the II-1, II-2, II-3, and II-4 probes are shown. The arrow shows divergent octamer-binding sequence in the II-4 region. B, substitution mutations were introduced into the II-1, II-3, and II-4 subregions of the 3.2nlacZ construct to generate construct 3.2OE1-IImnlacZ. The nucleotides that are not conserved in the human and mouse OE1 sequences are depicted in white uppercase letters. C, the expression of the lacZ reporter gene was increased in the ARC of the 3.2OE1-IImn-lacZ transgenic mouse. No ␤-gal ϩ /orexin-ir ϩ neurons were observed in the LHA. ␤-Galϩ, number of founder mice showing ␤-gal ϩ staining; Tgϩ, total number of transgenic founder mice.

TABLE IV
Quantitative reporter expression analysis (in transgenic founder mice) of a series of deletion constructs of the OE1 core region Transgenic founder mice were used for this analysis. Percentages of the ␤-gal ϩ population in orexin-ir ϩ neurons in the LHA and of ␤-gal ϩ neurons distributed in the medial regions of the hypothalamus (HT) were determined as described in the legend to Table I.

Construct
Transgenic mouse name ␤-gal ϩ orexin-ir ϩ ␤-gal ϩ in medial HT/ total ␤-gal ϩ in HT Multiple Substitution Mutations in the OE1 Core Region Abolish Prepro-orexin Gene Expression in the LHA-To investigate cis-acting motifs in the OE1 core region, we performed EMSA using nuclear extract from adult mouse hypothalamus. Using four oligonucleotide probes corresponding to the OE1 II-1 to II-4 sequences for EMSA ( Fig. 6A; see also Fig. 5A), we found that DNA-protein complexes were specifically formed with the II-1, II-3, and II-4 probes and could be effectively self-competed (data not shown). In contrast, no specific complex was formed with the II-2 probe. Thus, we conclude that the OE1 core region contains multiple binding sites for transcription factors.
Upon inspection of the OE1 core sequence, we did not find any significant motifs homologous to known consensus cisacting motifs for transcription factor binding, except one divergent Oct-1-binding site (25) in the OE1 II-4 zone (Fig. 6A). We also noticed that 48 of 58 nucleotides within the OE1 core were conserved between human and mouse sequences; divergent nucleotides are shown in white uppercase letters (Fig. 6B). Based on these lines of evidence and the results of additional EMSA analyses (data not shown), we hypothesized that the three motifs in the II-1, II-3, and II-4 subregions may be crucial for OE1 activity.
To further delineate the activity of the OE1 core region in regulating prepro-orexin gene expression in the LHA of transgenic mice, we generated the construct 3.2OE1-IImnlacZ by introducing substitution mutations in the II-1, II-3, and II-4 subregions of 3.2nlacZ (Fig. 6B). These mutations were selected based on similarity to known cis-acting motifs (II-4) or homology to mouse sequence (II-1 and II-2). Each mutation effectively eliminated the DNA-protein complexes in EMSA (data not shown). Examination of multiple transgenic mouse lines harboring 3.2OE1-IImnlacZ unequivocally demonstrated that the orexin neuron-specific expression of the lacZ reporter gene in the LHA was completely abolished by such mutations (Fig. 6C). On the other hand, the expression of the ␤-gal reporter clearly increased in the ARC (Fig. 6C), indicating that the II-1, II-3, and II-4 sequences within the OE1 core region are essential for regulating prepro-orexin gene expression in the LHA.

DISCUSSION
In this study, we analyzed the regulatory mechanisms governing human prepro-orexin gene expression using a reporter gene transgenic mouse assay. Because the prepro-orexin gene is expressed specifically in the lateral region of the hypothalamus, we investigated the precise region of the gene that determines its spatial expression profile. We found that a 214-bp OE1 region, located 287 bp upstream of the human preproorexin gene, is indispensable for expression of the reporter gene in orexin-ir ϩ neurons in the LHA. To our surprise, the construct delt-OE1nlacZ, which lacks OE1, did not give rise to reporter gene expression in orexin neurons, but instead directed expression ectopically in the medial regions of the hypothalamus, particularly the dorsomedial hypothalamus, the ventromedial hypothalamus, and the ARC. Based on these and other results, we conclude that the function of OE1 is to activate prepro-orexin gene expression in the LHA and to repress it in the medial regions of the hypothalamus.
Because the lacZ reporter gene expression observed in the medial regions of the hypothalamus was reproducible in three independent transgenic founder mice carrying delt-OE1nlacZ, the position effect variegation inherent to the transgenic mouse assay is unlikely to play a role in the expression site change. Rather, we assumed the presence of additional regulatory region(s) within the 3.2-kb region of the prepro-orexin gene to be directing gene expression in the medial regions of the hypo-thalamus. In this scenario, the prepro-orexin gene activity in the medial regions of the hypothalamus is normally either repressed or modified by OE1. Although OE1 activity directs reporter gene expression specifically in the LHA, mutations in OE1 shift the expression to the medial regions of the hypothalamus. In relation to this, the pro-opiomelanocortin gene, encoding the common precursor of ␣-melanocyte-stimulating hormone, ACTH, and ␤-endorphin, contains regulatory sequences specifying medial hypothalamic gene expression, namely in the ARC (26). The region 13 to 2 kb upstream of the pro-opiomelanocortin gene directs reporter gene expression in pro-opiomelanocortin neurons of the ARC (27). In preliminary experiments, we found that a population of ␤-gal ϩ neurons co-localized with ␣-melanocyte-stimulating hormone immunoreactivity in the ARC of the delt-OE1nlacZ transgenic mouse. 4 Thus, it is conceivable that a common gene regulatory mechanism exists in the pro-opiomelanocortin and prepro-orexin genes.
Only up to 50% of orexin-ir ϩ neurons in the 3.2nlacZ transgenic mouse were stained by ␤-gal histochemistry. Although the most straightforward interpretation of this observation is to assume the lack of an essential regulatory element in the 3.2nlacZ construct, an alternative interpretation is also possible and intriguing, in that this may also be due to the nature of the enhancer activity. Our current view of the enhancer function is that the enhancer acts to temporally open the chromatin structure surrounding the basal promoter region so that the gene can be transcribed during this period. Importantly, this opening period does not last long; and therefore, if the reporter protein is turned over rapidly, we cannot detect an entire set of orexin neurons as being 100% positive for reporter gene expression. Indeed, when we generated two lines of a transgenic mouse carrying the 3.2nEGFP construct, in which 3.2nlacZ contains enhanced green fluorescent protein in place of the lacZ reporter, we found that Ͼ80% of the orexin-ir ϩ neurons showed positive green fluorescence. 5 Because green fluorescent protein is known to be much more stable than ␤-gal, the discrepancy (50% of orexin-ir ϩ neurons are positive for ␤-gal, whereas 80% are positive for green fluorescent protein) may reflect reporter stability (28).
OE1 and the additional 0.7-kb region between 3.2 and 2.5 kb upstream of the prepro-orexin gene contain sufficient information for restoring gene expression in orexin neurons in the LHA. Within this 0.7-kb region, OE2 is likely to represent the critical regulatory sequence required, in tight cooperation with OE1, to regulate prepro-orexin gene expression specifically in the LHA. However, the repetitive Alu sequences are dispensable in targeting such spatial regulation. Indeed, the number of ␤-gal ϩ /orexin-ir ϩ neurons in the construct lacking OE2 (1.3nlacZ) was significantly lower than that in the construct containing OE2 (3.2nlacZ). Furthermore, although ␤-gal expression was observed in the LHA, the OE2-deficient 1.3nlacZ and 2.1nlacZ constructs resulted in ectopic lacZ reporter expression in non-orexin neurons of the ARC. One plausible explanation for these findings is that OE1 and OE2 in combination play a dual role in spatial expression: activation of preproorexin gene expression in the LHA concomitant with repression of gene expression in the medial regions of the hypothalamus. Thus, a reduction in LHA-specific gene expression accompanied by a leak of expression in the medial regions of the hypothalamus is inevitable with a loss of OE2 function.
Dissection of the OE1 domain (Ϫ287 to Ϫ77 bp) allowed us to decipher the precise regulatory sequence governing specific expression in the LHA. The critical sequence within OE1 re-quired for specific expression in the LHA was narrowed down to Ϫ258 to Ϫ201 bp. This OE1 core region defines the fundamental function of OE1, i.e. expression of the prepro-orexin gene specifically in orexin neurons in the LHA. Loss of any quarter of the 57-bp OE1 core region markedly reduced reporter gene expression, whereas no single quarter could represent the entire function of this core domain. Instead, multiple regulatory modules are likely to contribute to the total function of the OE1 core region to activate prepro-orexin gene expression in the LHA and to repress it in the ARC. The cis-acting elements within the OE1 core region might orchestrate transcriptional regulation through formation of an active enhanceosome (29). In the case of the insulin gene, for example, multiple cis-acting modules present in the promoter contribute to its overall activity in directing gene expression specifically in pancreatic ␤-cells (30 -32).
We speculate that the transcription factors that bind specifically to the OE1 II-1, II-3, and II-4 sequences, which are critical for gene expression in orexin neurons, are differentially expressed in the LHA and the medial regions of the hypothalamus. OE1 complexes composed of different transcription factors may perform dual functions to activate or repress preproorexin gene expression in the LHA and the medial regions of the hypothalamus, respectively. Accordingly, distribution of these transcription factors in the hypothalamus is of great interest in exploring the molecular mechanisms of such dual gene regulation.
In this regard, we found that the OE1 II-4 sequence contains a divergent Oct-1-binding site. Many members of the POU family of transcription factors that are capable of binding the octamer sequence have been identified in brain (33)(34)(35)(36)(37)(38)(39). For instance, Oct-1 plays a critical role in regulating the gonadotropin-releasing hormone gene, which is exclusively expressed in a discrete population of neurons in the hypothalamus (37,38). Brn2 is known to play an essential role in the development of the neuronal lineage in the paraventricular and supraoptic nuclei of the hypothalamus (36,39). In disagreement with our expectation, however, the II-4 sequence-specific binding factor does not appear to be a canonical POU family transcription factor, as the consensus octamer motif could not compete with the II-4 complex in EMSA. 4 It is intriguing to note that, in Xenopus, orexin immunoreactivity was observed exclusively in the medial regions of the hypothalamus (40). This observation suggests that the Xenopus prepro-orexin gene does not contain a region corresponding to OE1 of the human gene. It is conceivable that OE1 arose through evolution of the mammalian LHA to direct preproorexin gene expression specifically in this area. In conclusion, we have identified two transcriptional regulatory core sequences, OE1 and OE2, which, in combination, activate and repress gene expression in the LHA and the medial regions of the hypothalamus, respectively. The role that OE1 plays might explain the difference between the transcriptional regulatory mechanisms controlling LHA-specific prepro-orexin gene expression and the expression of other neuropeptide genes specific to the medial regions of the hypothalamus. Only a few examples of regulatory sequences capable of directing transgene expression in a neuron-specific manner are available. Thus, our study is an important step toward the elucidation of the regulatory mechanisms governing gene expression of the various medial and lateral hypothalamic neuropeptides and the mechanisms underlying neuronal differentiation during development of the hypothalamus.