Identification of the Downstream Targets of SIM1 and ARNT2, a Pair of Transcription Factors Essential for Neuroendocrine Cell Differentiation*

SIM1 and ARNT2 are two basic helix-loop-helix/PAS (Per-Arnt-Sim) transcription factors that control the differentiation of neuroendocrine lineages in the mouse hypothalamus. Heterozygous Sim1 mice also develop early onset obesity, possibly due to hypodevelopment of the hypothalamus. Although SIM1 and ARNT2 form heterodimers to direct the same molecular pathway, knowledge of this pathway is limited. To facilitate the identification of their downstream genes, we combined an inducible gene expression system in a neuronal cell line with microarray analysis to screen for their transcriptional targets. This method identified 268 potential target genes of SIM1/ARNT2 that displayed >1.7-fold induced expression. 15 of these genes were subjected to Northern analysis, and a high percentage of them were confirmed to be up-regulated. In vivo, several of these genes showed neuroendocrine hypothalamic expression correlating with that of Sim1. Furthermore, we found that expression of two of these potential targets, the Jak2 and thyroid hormone receptor β2 genes, was lost in the neuroendocrine hypothalamus of the Sim1 mutant. The expression and predicted functions of many of these genes provide new insight into both the Sim1/Arnt2 action in neuroendocrine hypothalamus development and the molecular basis for the Sim1 haplo-insufficient obesity phenotype.

VP neurons project to the posterior pituitary, where they release hormones directly into the bloodstream, whereas the CRH, TRH, and somatostatin neurons project to the median eminence, which in turn carries their secreted hormones to the anterior pituitary to modulate pituitary secretion (1). Despite extensive studies of the physiological functions of these hormones, the molecular pathways directing their expression in specific cell lineages are less well known.
Analyses of Sim1 and Arnt2 mutant mice have demonstrated that these transcription factors are essential for the terminal differentiation of the aforementioned neurons (2)(3)(4)(5). In the absence of either gene, the precursors of these neurons are born normally, but fail to form the anatomical neuroendocrine centers, i.e. the paraventricular nucleus (PVN) and the supraoptic nucleus (SON) in the anterior hypothalamus, and do not produce any of the hormones (2)(3)(4)(5). The collective loss of these neuroendocrine hormones may cause the observed perinatal lethality of the Sim1 and Arnt2 mutants (2)(3)(4)(5). Intriguingly, heterozygous Sim1 mice develop early onset obesity, proposed to be due to hypodevelopment of the neuroendocrine hypothalamus (6). A balanced chromosomal translocation disrupting SIM1 (7) and a haploid interstitial deletion of chromosome 6 encompassing SIM1 (8) have also been shown to be associated with profound obesity in humans.
Sim1 and Arnt2 are homologs of Drosophila sim and tango, respectively (9 -14). These genes belong to the family of basic helix-loop-helix (bHLH)/PAS (Per-Arnt-Sim) domain-containing proteins, many of which are important regulators of development and physiology (14). DNA binding assays in vitro have demonstrated that SIM1 and ARNT2 form heterodimers and bind the core sequence TACGTC, named central nervous system midline enhancer (CME) (3, 12, 14 -16). The CME was originally identified in the enhancer regions of sim/tango downstream genes (13,15). Multimerized CME can mediate sim/ tango-dependent central nervous system midline expression in the fly (11,(13)(14)(15)(16). When linked to a minimal adenovirus major late promoter-driven reporter gene, CME can also mediate SIM1/ARNT2-dependent transcriptional activation of the reporter in cultured mammalian cells, albeit weakly (17). Deletion analyses of SIM1 and ARNT2 demonstrate that their basic domains are required for CME recognition, their bHLH/PAS domains for heterodimerization, and their C termini for transcriptional regulation (9,12,17,18). However, when the SIM1 C terminus is fused to the Gal4 DNA-binding domain and tested in a different cell line using a Gal4-thymidine kinase promoter-driven reporter, it acts as a repressor (9). These results suggest that SIM1 can act as a repressor or an activator depending on the context of the reporter assay.
Brn2, a POU domain-encoding gene, is a downstream target of SIM1/ARNT2 in vivo (2)(3)(4)(5). In both Sim1 and Arnt2 mutants, Brn2 expression in the prospective neuroendocrine cells is lost. Furthermore, Brn2 mutant mice have a selective defect in CRH-, VP-, and OT-expressing neurons (19 -21), which is a subset of the Sim1 and Arnt2 mutant defects. BRN2 has also been shown to bind to the CRH promoter and to activate its transcription (22,23). The genes employed by SIM1/ARNT2 to specify the other neuroendocrine hormone gene expression in distinctive cell types remain unexplored.
To study the molecular pathways by which Sim1 and Arnt2 control the development of the hypothalamic secretory neurons and mediate energy homeostasis, we combined an inducible gene expression system with microarray analysis to screen for their downstream targets. Below, we describe the genes identified by this screen and the resulting implications for the Sim1/Arnt2-operated molecular pathway.
Cell Culture-The Neuro-2a cells (American Type Culture Collection, Manassas, VA) were cultured in Eagle's minimal essential medium (Vitacell, American Type Culture Collection) and 10% bovine serum. FuGENE 6 reagent (Roche Applied Science) was used for DNA transfection. For transient transfections, each plasmid was used at 250 ng in a final 1 g of total DNA for each well of a 6-well dish (Falcon). The plasmids used for each transfection are indicated in the figures. For a stable cell line, Neuro-2a cells were transfected with 10 g of pTet-On and selected with 200 g/ml G418 (Sigma) to obtain individual clones. Selected colonies were propagated and transfected with pTREhyg-Luc (Clontech) to test their Dox (1.5 g/ml; Clontech) responsiveness by assaying inducible luciferase activity. The clone with the lowest background was transfected with 2 g of pTRE-SN-VP16 and 10 g of pTet-tTS and selected with 150 g/ml hygromycin (Roche Applied Science) to obtain secondary clones. Individual clones were then tested for their Dox-regulated SIMN-VP16/ARNT2 activity by assaying for pML/ 6C-WT reporter activity under mock and Dox treatment conditions. Clone 37 was chosen for microarray study. Luciferase activities were measured by luciferin (Sigma) light emission using Monolight 2010 (Analytical Luminescence Laboratory).
Microarray Hybridization and Data Analysis-Total RNA was isolated using TRIzol reagent (Invitrogen), followed by the QIAGEN RNeasy method. Microarray hybridization using the MG-U74v2 A gene chip was performed using a service provided by Neurologic Functional Genomics. Triplicate hybridizations with independently synthesized probes were conducted using the same batch of RNA isolated from the untreated and Dox-treated clone 37 cell lines and the parental cell line. Hybridization signals were normalized by Affymetrix Suite software. The data sets were filtered by absent and present calls using the Affymetrix Datamining tool, i.e. genes that displayed inconsistent signals within the same oligonucleotide probe set are excluded. Based on the general background of the data, genes displayed signals Ͻ30 arbitrary units under both mock and Dox treatment conditions (after normalization) were excluded for further analyses. These remaining genes were subjected to Student's t test, with p Յ 0.05 considered significant. We arbitrarily chose a 1.7-fold increase as a cutoff threshold for selecting genes for further investigation based on Northern confirmation rate of genes displaying various fold inductions.
Cell lysates were resolved on a 4ϳ15% gradient gel (Bio-Rad); transferred to a Hybond membrane (Amersham Biosciences); and probed with antibodies against the VP16 activation domain (Clontech), ARNT2 (Santa Cruz Biotechnology), and ␥-tubulin (Sigma). Horseradish peroxidase-conjugated secondary antibodies (Sigma) coupled with the chemiluminescence reaction (Amersham Biosciences) were used to visualize these proteins.
In Situ Hybridization (ISH)-The brains of CD1 mice at embryonic day 18.5 (E18.5; vaginal plug date is designated as embryonic day 0.5) were snap-frozen in OCT compound and cryosectioned at 20-m thickness. In Fig. 6, Sim1 heterozygotes in a BL6 backcrossed background were mated to obtain E18.5 brains from Sim1 mutant and wild-type siblings for ISH. Digoxigenin-labeled sense and antisense probes of the genes specified in the figures were synthesized using SP6, T7, or T3 polymerase and used at ϳ1 g/ml for hybridization of the brain sections according to Schaeren-Wuemers and Gerfin-Moser (24).

Inducible Expression of SIM1 and ARNT2 in the Neuro-2a
Cell Line-Overexpression of transcription factors should lead to changes in the expression levels of their downstream target genes. To achieve controlled SIM1/ARNT2 overexpression, we chose to use the Tet-On inducible system (see "Experimental Procedures" for details). In this system, pTet-On and pTet-tTS ( Fig. 1A) are used to confer tetracycline-dependent regulation of pTRE (tetracycline-responsive element)-driven genes, in this case, Sim1 and Arnt2 ( Fig. 1A; diagrammed in Fig. 3A).
We reasoned to implement this system in a neuronal cell line, as we are interested in the function of Sim1/Arnt2 in the hypothalamus. Because the SIM1 C terminus has been reported to repress or activate transcription in different contexts (9,17), we surveyed the activity of SIM1/ARNT2 in various neuronal cell lines using the CME-driven luciferase reporter (pML/6C-WT) assay devised by Moffett and Pelletier (17). SIM1 and ARNT2 together activated this reporter expression in NB41A3, N1E-115, and Neuro-2a cells in a CME-dependent manner (data not shown). The Neuro-2a cells were chosen for further study due to their homogeneous morphology and high transfection efficiency.
Under the various Dox treatment conditions tested, we observed an optimal ϳ4-fold activation of the pML/6C-WT reporter by SIM1 and ARNT2 in Neuro-2a cells (Fig. 1B), regardless of whether they were expressed from a single plasmid linked by the IRES or from separate plasmids (data not shown). Concerned that their activity might be too low to activate endogenous genes for our assay, we constructed a potent SIM1 fusion activator. Sim1 hybrid constructs with the SIM1 C terminus replaced by the Gal4 and VP16 (SIMN-VP16) activation domains were cloned into the pTRE2hyg vectors pTRE-SN-Gal4 and pTRE-SN-VP16, respectively (Fig. 1A). Arnt2 was placed downstream of the IRES located 3Ј to these Sim1 variants for coexpression (Fig. 1A). Plasmids carrying these Sim1 variants as well as Arnt2 were transfected into Neuro-2a cells to compare their Dox-regulated proficiency in activating the pML/6C-WT reporter. Upon Dox treatment, pTRE-SN-VP16 conferred a 27-fold activation of the reporter compared with the 12-and 4-fold reporter activation rendered by pTRE-SN-Gal4 and pTRE-SF, respectively (Fig. 1B). SIMN-VP16 appears to function with similar specificity as SIM1, as it also required ARNT2 and wild-type CME sites for reporter gene activation (Fig. 1C).
pTRE-SN-VP16 was therefore used to establish a stable Neuro-2a clonal cell line for Dox-inducible expression of SIMN-VP16 and ARNT2 (see "Experimental Procedures" for details). One clone (clone 37) with this characteristic was obtained. As shown in Fig. 2A, this clone expressed readily detectable levels of the SimN-VP16-IRES-Arnt2 transcript of the predicted size of 4.2 kb upon Dox treatment. The SIMN-VP16 and ARNT2 proteins were also detected (Fig. 2B). No SimN-VP16-IRES-Arnt2 transcripts or SIMN-VP16 and ARNT2 proteins were megalovirus-derived enhancer/promoter for high level transcription; PminCMV, minimal cytomegalovirus-derived enhancer/promoter. B, comparison of the activities of SIM1 variants in the Tet-On system. Transient transfections of the Neuro-2a cells were carried out using the plasmids indicated. The fold induction was calculated as relative luciferase activities of the doxycycline-treated (Doxϩ) over the mock-treated (DoxϪ) samples. The relative luciferase activities were normalized to the ␤-galactosidase activity from cotransfected pSV-␤gal (data not shown). Results represent the mean of three independent experiments, and the error bars are the S.D. values. C, activation of the reporter by the SimN-VP16 fusion gene depends on Arnt2 expression from either a separate plasmid or a single plasmid linked by IRES and the wild-type CME, but not the mutant CME. Transfections, luciferase measurement, and normalization were the same as described for B. Arbitrary luciferase values were used to indicate CME activation.

FIG. 1.
A, schematic representation of the plasmids used in this study: pTRE2hyg, empty vector without a cDNA insert used as a control; pTRE-SF, pTRE2hyg harboring full-length Sim1 and Arnt2 joined by the IRES; pTRE-SN-Gal4, pTRE2hyg harboring the Sim1 N-terminal bHLH/PAS domain (SimN) fused to the Gal4 activation domain, IRES, and Arnt2; pTRE-SN-VP16, pTRE2hyg harboring SimN fused to the VP16 activation domain, IRES, and Arnt2; pTRE-Arnt2 and pTRE-Sim-VP16, pTRE2hyg harboring Arnt2 and SimN fused to VP16, respectively; pTet-On, for expression of the tetracycline-controlled transactivator rtTA; pTet-tTS, for expression of the tetracycline-controlled silencer tTS; pML/6C-WT (17), a luciferase reporter driven by the adenovirus major late promoter linked to four copies of CME (4ϫCME) for SIM1/ARNT2 binding; pML/6C-AM (17), a similar reporter linked to four copies of CME with point mutations (CMEm) as shown. TRE, tetracycline-responsive element for rTA and tTS binding; PCMV, cyto- A, experimental schemes of microarray screen for SimN-VP16/Arnt2 downstream targets using clone 37 cells. In the absence of Dox (DoxϪ; mock-induced), the tetracycline-controlled transcriptional silencer tTS bound to the tetracycline-responsive element (TRE) and repressed the transcription of SimN-VP16-IRES-Arnt2. In the presence of Dox (black dot, Doxϩ; Dox-induced), tTS dissociated from the tetracycline-responsive element, whereas the tetracycline-controlled transactivator rtTA bound to the tetracycline-responsive element and activated the transcription of SimN-VP16-IRES-Arnt2. Total RNA from each condition was isolated, labeled, and hybridized to microarrays, and the data were compared for differentially expressed genes under the two conditions. B, scatter plot of the entire set of gene expression data from the microarray. Each dot represents one gene. The normalized average signal intensity under the mock and Dox treatment conditions are shown on the x and y axes in arbitrary fluorescence values. Gray lines indicate the 1.7-, 2-, 10-, and 20-fold differential expression thresholds. The locations of three genes are marked: Arnt2, Jak2, and TR␤ 2 . Up-regulation of Arnt2 is an indicator that the Dox-treated sample did contain induced SimN-VP16-IRES-Arnt2 transcripts. C, table illustrating the distribution of the 268 up-regulated genes into seven categories. The transcription regulators represent genes that are known to participate in regulating transcription directly. The signaling component category includes secreted peptide factors, membrane receptors, and signaling effectors. Metabolic enzymes include enzymes in known metabolic pathways. Channels and transporters include neurotransmitter receptors or transporters for small molecules. The cell adhesion and migration category includes adhesion molecules, extracellular matrix proteins, and guidance molecules for cell migration and axonal projection. The miscellaneous group includes genes with documented functions not overlapping with the above categories, e.g. immunity-related proteins. The uncharacterized genes are expressed sequence tags in the data base whose functions have not been investigated. The functional categories and the number of genes within each category are indicated. tion). The presence of the Arnt2 gene on this chip serves as an internal control of Dox induction. A total of 268 genes displaying Ͼ1.7-fold (arbitrarily chosen) increased expression were considered significantly up-regulated by t test, with p Յ 0.05. These 268 genes can be divided into several functional categories as summarized in Fig. 3C and described in the legend thereof. The identity of each gene and its fold induction are listed in Table I. Importantly, no significant gene expression changes were found in the parental cell lines treated with Dox versus the mock-treated clone 37 cell line, indicating that the up-regulated genes in the Dox-treated clone 37 sample are regulated by SIMN-VP16/ARNT2.
Of note, we found that many of the up-regulated signaling component genes have been implicated in energy homeostasis (Table I), e.g. the MC3R, Jak2, and TR␤ 2 genes. Their possible roles in the Sim1/Arnt2-operated pathway will be discussed.
Independent Confirmation of the Microarray Results by Northern Analysis-We selected a group of target genes predicted by microarray analysis to perform Northern hybridization for independent confirmation. 15 genes were selected based on their possible relevance to the neuroendocrine system or obesity pathway (see Fig. 4 legend for the gene names). Total RNA samples of the mock-and Dox-treated clone 37 cells were subjected to Northern analysis using specific DNA fragments of these genes. Of the 15 genes, 12 were confirmed to be upregulated upon Dox treatment. Fig. 4 shows 9 of the 12 confirmed examples. The other three were not confirmed, as they did not give a detectable signal (data not shown). Of note, the fold changes observed by Northern analysis were generally higher than those observed by microarray analysis.
Expression of Potential Sim1/Arnt2 Downstream Targets in the Neuroendocrine Hypothalamus-As a secondary screen, we used ISH to examine whether some of these potential target genes have expression domains overlapping with the Sim1positive domains in the hypothalamus. Five genes, Jak2 and IL-6R␣ (of the signaling component category), TR␤ 2 and Tlx2 (of the transcription regulator category), and Chrne (of the neurotransmitter receptor category), were chosen for this survey. ISH was carried out on sectioned E18.5 mouse brains using digoxigenin-labeled antisense probes of each gene. Adjacent sections throughout each brain were hybridized to the Sim1 probe and the gene probe of interest for comparison. The Sim1-positive regions are diagrammed in Fig. 5A (based on the data in Fig. 5B). In addition to its expression in the PVN and SON (10), Sim1 expression was also found in the nucleus of the lateral olfactory track (NLOT) in the amygdala and in the anteroventral nucleus (AVN) in the hypothalamus (Fig. 5B).
When Jak2 expression was compared with that of Sim1 in adjacent sections, we noticed that Jak2 was expressed at the ventral PVN (Fig. 5D), occupying a subdomain of the Sim1 domain (Fig. 5C). Expression of Jak2 was localized to the mid-to-posterior level of the PVN. On the other hand, TR␤ 2 gene expression was found in the dorsal region of the PVN Sim1 domain at the posterior PVN level (Fig. 5, compare G and  H). Expression of the IL-6R␣, Tlx2, and Chrne genes was also found at the mid-to-posterior PVN, and each displayed a distinctive subdomain expression pattern within the PVN Sim1 domain (Fig. 5, compare K-N). With the exception of Tlx2 (Fig.  5Q), expression of the other four potential Sim1 target genes was also found in the SON (Fig. 5, E, F, I, J, O, P, and R). All of these genes were also commonly expressed in the AVN and NLOT (Fig. 5, C-R). Sense probes of these genes gave no specific signals in the regions of interest: the PVN, (Fig. 5, CЈ, DЈ, GЈ, HЈ, LЈ, MЈ, and NЈ) and SON and NLOT (Supplemental Fig. 1). Thus, of the five genes surveyed, all displayed expression domains overlapping with those of Sim1.
Expression of the Jak2 and TR␤ 2 Genes Is Down-regulated in the Sim1 Mutant-If these identified potential targets are true downstream genes of Sim1/Arnt2, we reasoned that their expression in Sim1 mutants should be significantly reduced or lost. We chose to examine Jak2 and TR␤ 2 gene expression in the Sim1 mutant by ISH. These two genes were chosen because of their potential relevance for hypothalamus-mediated energy homeostasis. Importantly, Sim1/Arnt2 is essential for PVN cell terminal differentiation, but not for cell genesis or survival. Sim1-positive cells are found in the Arnt2 mutant located laterally to the normal PVN position (3). In addition, in a lacZ knock-in allele of Sim1, LacZ-positive mutant cells were present in the same ventral location as the Sim1-positive cells in the Arnt2 mutant. 2 In contrast to its expression in the wildtype PVN and AVN, Jak2 expression in the Sim1 mutant was greatly decreased in the presumptive PVN and AVN area (Fig.  6, compare A and B). Similarly, TR␤ 2 gene expression was greatly reduced in the presumptive PVN and AVN area in the Sim1 mutant (Fig. 6, compare C and D). This set of data suggests that Sim1 (and by inference, Arnt2) acts upstream of the Jak2 and TR␤ 2 genes.

From a Neuronal Cell Line to the Neuroendocrine System-
Here, we used the Tet-On system in Neuro-2a cells to overexpress SIMN-VP16/ARNT2 and screened for potential downstream genes of Sim1/Arnt2 in the neuroendocrine hypothalamus. An ecdysone-inducible system was also used recently in the neuronal SHEP cell line to obtain downstream genes of NPAS2/BMAL1 in the cortex (25). Together, these two examples lend support to the general applicability of such an experimental design.
There are, however, limitations to this approach. First, chromatin accessibility of certain target genes may be limited in a given cell line. Second, the cell lines used may not provide tissue-specific accessory factors needed for SIM1/ARNT2 function. Third, there may be fortuitous activation of genes that contain CME in their promoters, but that are neither coexpressed with nor regulated by Sim1/Arnt2 in vivo. Although the first issue cannot easily be overcome, we minimized the other concerns by surveying various neuronal cell lines, modifying SIM1 by fusing it to the VP16 activation domain, and performing ISH as a secondary screen for in vivo coexpression and regulation.
Comparison of Microarray and Northern Data-Our results and those of others (26) demonstrated that the fold differences measured by Northern analysis are generally higher than those measured by microarray. Such a discrepancy may be due to inherent differences in these two methods. Yet, this observation raises the possibility that target genes displaying low threshold differences are unavoidably excluded by applying stringent statistical criteria to the microarray data. On the other hand, by choosing a 1.7-fold increase as a cutoff value for up-regulated genes, we found that 12 of the 15 genes tested that gave signals on the Northern blots are up-regulated. For the three genes that did not give signals, they may be false predictions, or they are in fact up-regulated, but expressed at too low a level for detection by Northern hybridization using total RNA. Based on these data, we expect that of the 268 genes, a similar percentage of them can also be confirmed independently.
Hypothalamic Expression of the Potential Downstream Genes-Several up-regulated genes selected for ISH show expression patterns overlapping with the Sim1 expression pattern. Specifically, these selected genes are expressed in discrete    (27,28). To our surprise, all five genes are also commonly expressed in the AVN and NLOT, where Sim1 is also expressed. Although the physiological functions of these two nuclei are unknown, our data suggest that they are developmentally or functionally linked to the PVN and SON. In addition, the extensive correlation of expression between these genes and Sim1 strongly argues for their regulation by Sim1 in vivo. Regulation of Hormone Genes by Sim1 and Arnt2-Although the oligonucleotides representing the OT, VP, and TRH genes are on the chip, they are not found in the up-regulated gene group. This is likely due to the fact that they are specialized markers for mature secretory neurons, and their chromatin organizations are not accessible in Neuro-2a cells. We are, however, surprised that Brn2 was not up-regulated by both microarray and Northern analyses (data not shown). Shh was also not in the up-regulated pool (data not shown), even though Sim1 overexpression activates ectopic Shh expression in transgenic mouse embryos (29). The lack of Shh and Brn2 in the up-regulated pool suggests that, although our approach is fruitful, it does not uncover all in vivo targets.
Among the five hormone genes (VP, OT, CRH, TRH, and somatostatin) whose expression is missing in Sim1 and Arnt2 mutants, none has been shown to be a direct target of Sim1/ Arnt2 (2)(3)(4)(5)19). It is likely that Sim1 and Arnt2 govern a second tier of transcription factors, which in turn act independently or in combination to direct each hormone gene expression. For example, Brn2 acts downstream of Sim1/Arnt2, and its gene product can bind and activate the CRH promoter (22). On the other hand, the TRH promoter can be activated by TR␤ 2 in a thyroid hormone-independent manner in cultured cells (30,31). Our finding that the TR␤ 2 gene is a target of Sim1/Arnt2 suggests that Sim1/Arnt2 utilizes the TR␤ 2 gene to activate the TRH gene during development. In addition, our screen uncovered 38 potential Sim1/Arnt2 targets in the transcription factor category. They may be employed by Sim1/Arnt2 to orchestrate the expression of lineage-specific hormone genes such as OT, VP, and somatostatin.
Genes Involved in the Energy Homeostasis Pathway-Sim1 heterozygotes develop early onset obesity. It has been proposed that PVN hypodevelopment of embryonic origin instead of PVN malfunction of adult origin is the contributing cause (6). We were therefore surprised to find that several genes directly implicated in the negative regulation of body weight (but not in PVN development) turn up in the up-regulated pool, including the MC3R, Jak2, TR␤ 2 , and IL-6R␣ genes. The MC3R gene mediates the melanocortin-stimulating hormone signaling pathway, and MC3R mutant mice develop obesity (32,33). Jak2 is thought to be a positive participant in leptin receptor signaling (34,35). Both melanocortin-stimulating hormone and leptin are well known negative regulators of food intake (36). Although their actions have been investigated in the arcuate nucleus, their actions in the PVN are less well studied. The TR␤ 2 gene, being a thyroid hormone nuclear receptor, is likely to participate in thyroid hormone-mediated metabolic rate increases. IL-6 mutant mice also develop obesity (37). Expression of the IL-6R␣ gene in the PVN suggests that the PVN may be one of its action sites. The fact that these four genes are expressed in the PVN raises the possibility that Sim1 positively regulates these genes in the adult PVN. The collective reduction in the expression levels of these genes in the PVN of Sim1 heterozygotes may lead to energy homeostasis imbalance and the obesity phenotype observed. Potential Target Genes of Other Functional Categories-There are many potential target genes of Sim1/Arnt2 in other functional categories. Genes such as Chrne, Grin1, and Homer1 are involved in neurotransmitter reception (38 -40). Their potential downstream status supports the possibility that Sim1/ Arnt2 acts to confer or regulate the neuronal connectivity of PVN and SON cells. The cell adhesion and migration category of genes also has important implications. As mentioned above, the presumptive PVN and SON cells in the Sim1 and Arnt2 mutants stay at a lateral position (with respect to the normal PVN) because they fail to congregate centrally and migrate laterally to form the anatomical PVN and SON, respectively (6). 2 Genes in this class may facilitate the structural formation of the PVN and SON as well as axonal projections of the neurons therein. Together, these two categories of genes may help to build the structural and functional PVN and SON.
It is puzzling that metabolic enzymes such as glucose-6phosphate dehydrogenase and NADH dehydrogenase are in the up-regulated pool. The function and tissue-restricted expression of Arnt2 and Sim1 do not support their general role in regulating metabolic enzyme levels. We suggest that these genes are targets of another bHLH/PAS protein pair(s) with DNA binding specificity similar to that of SIM1/ARNT2. One such candidate is the dimer hypoxia-inducible factor 1␣/ARNT, as it is known to regulate many metabolic enzyme genes (41).
Direct Downstream Targets of SIM1/ARNT2-There are many genes found in the down-regulated population. Because we used SIMN-VP16, a stronger activator, the down-regulated genes are likely a result of secondary events mediated by the induced transcriptional repressors. The same reasoning can also be applied to the up-regulated population, as some of the genes may also be activated by secondary transcription factors. For the Jak2 and TR␤ 2 genes, whose PVN expression is dependent on Sim1, we searched for the CME consensus sequence (ACGTG) near their loci. Within 10 kb upstream of the mouse Jak2 coding region, there are five CME sites, two of which are also found in similar locations in the human JAK2 locus. For the TR␤ 2 gene, there is one CME site located within the first 1 kb of the promoter region in both the human and mouse genes. Although this is consistent with their being direct targets of SIM1/ARNT2, extensive enhancer/promoter analysis will be necessary to reach a firm conclusion. Even so, the diverse and rich composition of the genes that we have uncovered in this study provides an advantageous tool to explore the developmental and physiological processes of the neuroendocrine system.