Regulation of Tryptophan Hydroxylase-2 Gene Expression by a Bipartite RE-1 Silencer of Transcription/Neuron restrictive Silencing Factor (REST/NRSF) Binding Motif*

Tryptophan hydroxylase-2 (TPH2) is the rate-limiting enzyme in raphe serotonin biosynthesis, and polymorphisms of TPH2 are implicated in vulnerability to psychiatric disorders. Dynamic transcription regulation of TPH2 may underlie differences in vulnerability. We identified a transcription element in the TPH2 promoter that resembles the binding motif for RE-1 silencer of transcription (REST; also known as NRSF) transcription factor. REST limits tissue expression of non-neuronal genes through a canonical 21-bp motif called the NRSE (neuron-restrictive silencing element). The NRSE in TPH2 is a novel bipartite variant interrupted by a 6-base insertion. We confirmed that this bipartite NRSE permits transcriptional repression by REST identical to canonical NRSE in rat C6-glioma cells. Synthetic permutations of the motif revealed considerable flexibility in the juxtaposition of the two halves of bipartite NRSE. Computational analysis revealed many bipartite NRSE variants conserved between mouse and human genomes. A subgroup of these was found to bind REST by chromatin immunoprecipitation. Messenger RNAs for TPH2 and potassium channel H6, another gene with a bipartite NRSE, were up-regulated by dominant-negative REST in C6-glioma cells. These findings, which indicate that TPH2 expression is part of the developmental program regulated by REST and suggest that many previously unrecognized genes may be regulated by REST through the novel motif, have implications for the mechanism of REST action.

expression in the brain is restricted to raphe neurons, and TPH2 is likely the gene coding for the majority of tryptophan hydroxylase in these cells (1,2).
Serotonin modulates a variety of complex behaviors (3), and its signaling is implicated in the mechanism for a subset of antidepressants (4), but its precise role in mood regulation is unclear. A number of manipulations, including restriction of the essential dietary precursor tryptophan (5), stimulant drugs of abuse (6), or adverse developmental experiences (7), result in low brain serotonin levels and may underlie associated depression or anxiety phenotypes. Functional variants of TPH2 in mice are associated with decreased brain serotonin levels (8), and a TPH2 polymorphism in humans arguably (9) confers vulnerability to major depression (10). Associations with TPH2 polymorphisms are also reported for suicide (11), obsessivecompulsive disorder (12), and attention deficit disorder (13)(14)(15).
In contrast to the growing number of genetic association studies, little is known about the regulation of TPH2 expression. Limited studies suggest that TPH2 mRNA levels are decreased following stress corticosteroids (16), unchanged (16) or increased by ovarian steroids (17), and increased by hypotensive stress (18) and one recent report suggests that POU3F2 may regulate TPH2 through a functional single-nucleotide polymorphism in the promoter (19). Here, we report a functional motif in the TPH2 promoter that is a novel variant of the binding site for REST (RE-1 silencing transcription factor), known alternatively as NRSF (neuron-restrictive silencing factor).
The NRSE-like element in TPH2 is evolutionarily conserved but distinguished from the 21-bp canonical NRSE by the presence of a 6-bp insertion in the middle of the motif. We demonstrate that this bipartite NRSE (bNRSE) functions as a repressor through REST and that additional functional, evolutionarily conserved bNRSEs with variable insertions are found throughout the genome. Our findings indicate TPH2 is transcriptionally regulated by REST through a novel bNRSE. Moreover, these results have implications for the mechanism of action and the potential transcriptome regulated by REST.

EXPERIMENTAL PROCEDURES
Clones-[SCN2a promoters lacking (Null) or including 113 bp of upstream sequence flanking the SCN2a NRSE (SCN113) were reengineered for luciferase reporter from clones kindly provided by Dr. Gail Mandel (31).] The Null vector contains rat SCN2a minimal promoter/5Ј-UT as a 325-bp BglII-HindIII fragment in pGL3Basic (Promega, Madison, WI). Mouse TPH2 promoter and 5Ј-UT sequences (607 bp) were PCR-cloned from bacterial artificial chromosome RPCI23 151J17 obtained from the Children's Hospital Oakland Research Institute (Oakland, CA) using the primers mT2-MluI and mT2-NcoI (primer sequences are available in supplemental Table S1 online). All other NRSE motifs and mutants were synthesized as DNA oligonucleotides engineered with compatible overhangs to permit directional cloning into the polylinker NheI and BglII sites upstream of SCN2a minimal promoter in Null. To test the TPH2 bNRSE in its native context, the mouse 607-bp MluI-NcoI-flanked TPH2 promoter amplicon described above was cloned between MluI and NcoI of GL3Basic. All clones were sequence verified. An expression vector for human REST (REEX1) and dominant-negative REST (Dn-REST; p73) were gifts from Dr. Gail Mandel (31). Trichostatin A (made 1 mM in ethanol) and sodium butyrate (made 1 M in water) were purchased from Sigma.
Cell Culture-C6-glioma cells were obtained from American Type Culture Collection (Manassas, VA) and maintained in the recommended media. All transfection studies were repeated at least three times. For a typical transfection 50,000 cells/well were plated in a 24-well plate and transfected in triplicates. Transfection complexes with Lipofectamine 2000 (Invitrogen) typically contained 0.3 g of promoter-reporter construct, 0.6 g of either control (empty vector), Dn-REST, or REST, and 0.1 g of a 1:1 mixture of pEGFP-C2 (Clontech, Mountain View, CA) to monitor transfection efficiencies by fluorescence and pCMV-Sport-Bgal (Invitrogen) for reference ␤-galactosidase expression. Trichostatin A (100 nM) or sodium butyrate (5 mM) was added 4 h after transfection. After 48 h, cells were washed with phosphate-buffered saline and harvested with 50 l of reporter lysis buffer, and 5 l was assayed in parallel for luciferase (Promega) and ␤-galactosidase using Galacton-Plus (Applied Biosystems, Foster City, CA). Relative luciferase values were corrected for ␤-galactosidase values. Each experiment was conducted at least three times, and normalized average results across experiments are reported. Puromycin, obtained from Fisher Scientific, was resuspended in water at 2 mg/ml. The puromycin resistance gene was subcloned under the human ubiquitinC promoter in pUS2 backbone (38). Two g/ml puromycin killed Ͼ95% of C6-glioma cells in 48 h. Sta-tistical analyses were conducted in Microsoft Excel or Graphpad Prism for Windows PC.
Computational Analysis-We searched a locally constructed and publicly available Oracle data base (arrayanalysis.mbni. med.umich.edu/blast/). We initiated the search in mouse (mm6; build 34) because the mouse genome browser in the University of California, Santa Cruz (UCSC) data base contains the most current multispecies alignments (39). Sequence output from a search for the 5Ј-half NRSE in this data base was processed with a PERL (www.activestate.com/) script that searched for a nearby 3Ј-half motif. Genomic locations for the resulting subset were uploaded to the UCSC mouse genome browser as a custom track, and an intersection table of precalculated "multiz10way" alignments was retrieved. Subsequent PERL scripts were used to parse the human and mouse tracks and limit the set to only those containing 5Јand 3Ј-half bNRSE matches in both species. To annotate gene locations, PERL scripts were used to compare all matches with the search spaced defined by annotated Ensemble genes (www.ensembl.org). Annotated results of this computational search are shown in supplemental Table S2. Where available, the flanking DNA sequence in the orthologous rat genome was retrieved for PCR primer design in Primer3 (40) (see supplemental Table S1 for primer sequences).
Chromatin Immunoprecipitation (ChIP)-ChIP experiments were performed essentially as described in the ChIP assay kit (Upstate, Waltham, MA). Briefly, ϳ5 ϫ 10 5 cells/antibody were fixed with 2.5% formaldehyde for 10 min at 37°C. Cells were washed extensively with phosphate-buffered saline, the lysate was cleared by centrifugation, and the chromatin was sheared by sonication to ϳ200 -1000-bp fragments. Samples were precleared with protein G-agarose/salmon sperm DNA slurry (Roche Applied Science). Antibodies to REST P-18 (Santa Cruz Biotechnology, Santa Cruz, CA) or FLAG M2 (Sigma) were added, and antibody/chromatin complexes were collected by incubation with protein G-agarose/salmon sperm DNA slurry. Samples were washed and eluted from resin as per the kit. PCR was performed with the indicated primer pairs (supplemental Table S1) and iQ Supermix (Bio-Rad) for 35 cycles at an anneal temperature of 57°C. For each result, two separate ChIP assays were run with identical results.
Quantitative Reverse Transcription-PCR-was conducted on a MyIQ cycler (Bio-Rad) using AmpliTaq/Syber-Green (Applied Biosystems) according to the manufacturer's instructions. First-strand cDNA template was made from 1 g of DNase-treated C6-glioma cell total RNA using iScript reverse transcriptase, oligo(dT)/random hexamer mix, buffer, and the reaction conditions specified in the iScript cDNA synthesis kit (Bio-Rad). Intron-spanning primers for hypoxanthine-guanine phosphoribosyltransferase (HPRT), TPH2, and KCNH6 (listed in supplemental Table S1) were verified to amplify a single amplicon of the correct size and sequence. All PCR reactions were run in triplicate, and the entire experiment was replicated. PCR conditions were 35 cycles of 95°C for 15 s, 57°C for 10 s, and 72°C for 15 s. 10-Fold serial dilutions of template were used in PCR reactions to estimate amplification efficiencies. Relative expression levels of TPH2 and KCNH6 were normalized to the expression of HPRT according to the method of Pfaffl (41).

RESULTS
We aligned human, chimp, dog, mouse, and rat TPH2 promoters, revealing a highly conserved 21-bp motif that matched the 5Ј-half of the consensus NRSE (42) but lacked identity with the 3Ј-half ("NRSE-like"). Closer inspection revealed a perfect match to the 3Ј-half 6 bp downstream relative to its expected location (Fig. 1A). This motif is located in the vicinity of the cloned mammalian TPH2 mRNA 5Ј-ends (Fig. 1B). To determine whether this NRSE variant would function as a REST silencing element, we engineered a luciferase expression vector for testing NRSE variants based on the previously published rat sodium channel 2a (SCN2a) promoter (20). A "Null" control contained the rat SCN2a minimal promoter. SCN113 contained the 113-bp SCN2a NRSE upstream fragment, and SCN21 contained only the core SCN2a NRSE in the upstream polylinker. SCN21 differs from the earlier report (31) in two ways: (a) it contains a luciferase reporter instead of chloramphenicol acetyltransferase, and (b) the NRSE is introduced as only the core 21-bp motif. A "scrambled" control (Scr21) contained a random 21-bp sequence instead of the 21-bp NRSE ( Fig. 2A). To verify our modified constructs, we tested repression activity in rat C6-glioma cells with or without Dn-REST, a dominant-negative inhibitor of REST (31). C6-glioma cells FIGURE 1. The rat TPH2 promoter and bipartite NRSE motif. A, conserved bNRSE in the TPH2 promoter. 5Ј-and 3Ј-half motifs (bold) aligned to a consensus NRSE in the SCN2a promoter. **, purine pair critical for NRSE function;^a nd ϩϩ, other dimers mutated in current studies. The last four nucleotides of the degenerate NRSEs are shown as nnnn to highlight their variable inclusion in published consensus motifs (42,43). B, rat TPH2 gene promoter and part of exon 1. Numbering is relative to the first nucleotide of the predicted translation start codon. The bNRSE motif is underlined, and the putative TATAAA motif for RNA polymerase II complex formation is bolded. TPH2 forward and reverse primers for ChIP analysis are labeled TPH2.fwd and TPH2.rev. The 5Ј-nucleotide of the rat cDNA (GenBank TM accession number NM_173839) has an R underneath it. Also labeled are the 5Ј-nucleotides of ClustalW-aligned orthologous cDNAs for human (H), mouse (M), Macaca fascicularis (f), and Macaca mulatta (m) TPH2 (GenBank TM accession numbers NM_173353, NM_173391, AB097528, and AY827483, respectively). were previously shown to express full-length REST and mediate repression of NRSE bearing heterologous promoters (34). The truncated version (SCN21) behaved identically to the longer SCN113 (20). Both exhibited ϳ90% repression in C6-glioma cells and reversed Ͼ65% in the presence of dominant-negative REST (Fig. 2B). The Null and scrambled controls exhibited similar levels of reporter.
We tested in this reporter the 21 bp "NRSE-like" motif (T2N21) from mouse TPH2 (identical in rat and human) as well as a longer variant including the complete 5Ј and 3Ј motifs (T2N27). T2N21 exhibited moderate repression activity in C6-glioma cells. However, this degree of repression (typically ϳ50% decrease) was much less than that seen with SCN21. T2N27, by contrast, repressed comparably to SCN21, and was comparably derepressed by Dn-REST (Fig. 2B). We call the extended motif a "bipartite NRSE" as distinguished from the "classic" 21-bp NRSE. Given that the 2 bp in the middle of classic NRSE are ambiguous, we refer to the sequence separating the 5Ј and 3Ј-halves as a "spacer," being 2 bp in classic NRSE and 8 bp in bNRSE of TPH2. T2N21, which contains 4 bp of the 3Ј-half, in fact, reconstituted 6 bp because of the adjacent "AG" of the BglII site and actually represented a bipartite version of the recently reported consensus motif (43), consistent with partial repression by this construct. Neither the 5Ј-half nor the 3Ј-half by itself was an effective repressor, although the 3Ј-half showed some mild repression (Fig. 3B).
REST represses or silences gene expression by several mechanisms, some of which involve recruitment of histone deacetylase. To determine whether the silencing activity observed is mediated by histone modification, we tested activity in the presence of trichostatin A, a histone deacetylase inhibitor. As shown in Fig. 3A, 100 nM trichostatin A partially reversed silencing of heterologous constructs containing classic and bNRSEs. As a percentage of basal expression, trichostatin A was equally effective (250 -300% increase) in derepressing promoter activity but had no effect on a construct lacking an NRSE. Five millimolar sodium butyrate, another inhibitor of histone deacetylase (44), had a comparable effect on the bNRSE motif (Fig. 3A).
T2N27 contains a duplication of the "RG" dinucleotide ( Fig.  1A, **) reported to be critical for NRSE function (45). One GG pair is topographically preserved with respect to the 5Ј-half NRSE (ϩϩ), and a second pair is preserved with respect to the 3Ј-half NRSE (**). We mutated each GG pair to TT separately or in concert in T2N27. Mutation of the first (ϩϩ) GG pair (T2N27m2) had no effect on silencing activity (Fig. 3B). Mutation of the second (**) GG pair (T2N27m3) markedly reversed repression. Mutation of both pairs (T2N27m2/3) had the same effect as T2N27m3. Near complete derepression was also achieved by a GC to TT mutation (T2N27m1) in the 5Ј-half motif (Fig. 1, [caret] [caret]). A double mutant containing both functional mutations (T2N27m1/3) was no different from each mutation separately (Fig. 3B).
To establish that REST is bound to bNRSE of TPH2 in vivo, we carried out ChIP assays in C6-glioma cells. REST antibody specifically pulled down chromatin fragments containing the SCN2a and the TPH2 promoter, but it did not pull down a fragment of the TPH2 gene 106 kilobases 3Ј to the promoter (Fig. 3C). Immunoprecipitates using a control antibody (anti-FLAG) lacked any of these amplicons. We conclude from these findings that REST binds the bNRSE motif in the TPH2 promoter and that the critical RG pair (45) constrains only the 3Ј-half of bNRSE-mediated silencing, but mutation of either half-site abolishes repression activity.
The bNRSE in TPH2 is located near the transcription initiation site predicted from cloned cDNAs from several mammals

. bNRSE in TPH2 promoter is partially reversible by trichostatin A in C6-glioma cells and binds REST in vivo.
A, repression by the TPH2 promoter bNRSE is partially reversed by trichostatin A or sodium butyrate and is comparable with that seen for classic NRSEs from the SCN2a and BDNF (BN21) promoters. *, p Ͻ 0.05; **, p Ͻ 0.01 compared with control condition by analysis of variance followed by Tukey's post hoc test. B, mutation of critical residues in the 5Ј-and 3Ј-halves abolishes repression (see "Experimental Procedures" for description of constructs). Relative expression, shown as the mean Ϯ S.E. of triplicate experiments is normalized to Scr21 in the absence of Dn-REST. **, p Ͻ 0.01 compared with Scr21 control by analysis of variance followed by Tukey's post hoc test. C, ChIP assay showing positive control amplicons from input DNA, amplicons encompassing the SCN2a NRSE and TPH2 bNRSE immunoprecipitated by ␣-REST antibody in C6-glioma cells that lack an amplicon by control antibody (␣-FLAG) or by a target lacking an NRSE (TPH2 exon 11). (Fig. 1B). To determine whether it is a repressor in its native context, we tested 607 bp of the promoter and 5Ј-UT of mouse TPH2 (through the ATG start codon) in pGL3-luciferase (clone T2N607). As with the SCN2a reporter constructs, we observed potent repression (ϳ75%) that was reversed by Dn-REST (Fig.  2B). Conversely, in mouse neuro-2A cells, which lack functional REST, neither the TPH2 bNRSE motif by itself nor the 607-bp TPH2 promoter was repressed, and in fact, they showed enhancer activity relative to scrambled control that was repressed by co-transfection with full-length REST (Fig. 2C).
To ask whether the 5Ј-and 3Ј-halves of bNRSE are topographically constrained, we tested constructs containing 2-bp step changes in the size of the spacer (Fig. 4A). The 21-bp canonical length (T2sp21) and T2N27 showed the most potent repression, and other sizes showed attenuated responses (Fig.  4B). To determine the orientation dependence, we tested variants in which the 5Ј-and 3Ј-halves of the NRSE were inverted or swapped. We tested these in the 21-bp version, in which is it unlikely that the DNA strand could loop back on itself to reorient "head-to-tail." When the 3Ј-half preceded the 5Ј-half, repression was completely lost, but variants in which the half motifs were oriented in a "head-to-head" or "tail-to-tail" fashion retained substantial repression activity (Fig. 4B).
To discover other genes that may be regulated by bNRSEs, we searched the mouse genome for all matches to the 5Ј-motif "TYAGMRCC" followed by a secondary search for the 3Ј-motif "RGMSAG" within 30 bp downstream. This initial search pro-duced 22,553 matches, revealing an excess of matches of classic length NRSE (spacer ϭ 2 bp) and a secondary peak at the size corresponding to the TPH2 bNRSE (spacer ϭ 8 bp). Further analysis was limited to the 12,276 matches with spacers of 11 bp or less. We observed approximately the same number of classic length NRSEs (1655) as reported previously (1894), but as noted therein, many are far from known genes, are located near genes lacking a restricted neuronal expression, or fail to bind REST (43). Because evolutionarily conserved motifs are more likely to be functionally significant, we focused on those bNRSEs conserved between human and mouse based on the precalculated human-mouse alignments in the UCSC browser (39). This conserved subset contains only 575 matches for spacers from 0 to 11 bp. Hence, only ϳ5% (575/12,276) of bNRSEs, based on the consensus motif employed, are conserved between mouse and human genomes. Within the subset of 575 conserved NRSEs, 258 are of the classic length (spacer ϭ 2). This group comprises all of the previously defined NRSEs including those for SCN2a, SCG10, and BDNF (supplemental Table S2). The distribution of conserved bNRSEs of other spacer lengths (Fig. 5) reveals a potential secondary peak at spacer ϳ8 bp. There were 53 matches with spacer ϭ 8 and 42 matches with spacer ϭ 9, as compared with 18 -29 matches for spacers of length other than 2, 8, or 9. Of the 575 matches, 310 (54%) are within a human Ensemble annotated gene, 400 (69%) are within Ϯ10 kb, and 434 (75%) are within Ϯ25 kb of annotated genes (supplemental Table S2).
Representative bNRSEs were selected for ChIP analysis based on evolutionary conservation and availability of the corresponding rat gene sequence for PCR primer design. We intentionally selected motifs with longer spacers, as found in the TPH2 variant, and noted that conserved motifs are biased for pyrimidines in the four nucleotides following the 3Ј-half degenerate search string, consistent with our observation that some of this sequence is necessary for potent repression (Fig.  2B). Many of the candidate bNRSE-containing amplicons tested were successfully enriched after ChIP with REST antibody (Fig. 5B). Representative classic NRSEs in BDNF and SCG10 also yielded expected amplicons.
We used quantitative reverse transcription-PCR to determine whether dominant-negative inhibition of REST could derepress TPH2 and KCNH6 (two of the genes with bNRSEs verified by ChIP). Dn-REST or control expression vector was co-transfected with a puromycin resistance expression vector in C6-glioma cells. Transfected cells were selected with puromycin and harvested at 72 h for cDNA template. Amplification efficiencies for the three primer pairs used were 1.98 for the reference gene HPRT (NCBI Entrez Gene ID: HPRT), 1.93 for TPH2, and 1.97 for KCNH6. The Ct (threshold cycle) difference for HPRT was less than 0.5, indicating that Dn-REST had no appreciable effect on reference gene expression. Based on two separate experiments, TPH2 and KCNH6 mRNAs were up-regulated by Dn-REST a minimum of 4.4-and 2.4-fold relative to HPRT, respectively (Fig. 5C).
Several features surfaced through examination of conserved bNRSE sequences. In some cases the gap sequence diverges (e.g. for DPYSL4), whereas in other cases (e.g. BAI3) it is highly conserved (Fig. 6A). Orthologous bNRSEs for some genes, e.g. DPYSL4, contain different spacer sizes. Some genes contain conserved perfect matches to only one-half of the NRSE. For example, it was recently reported that the SCN8A gene con-tains a perfect match to the 5Ј-half NRSE, but the flanking sequences, which are also conserved, differ markedly from the NRSE consensus (46). One possibility is that this motif assembles a distinct complex in which REST dimerizes with a different partner. Alternatively, and for which SCN8a may be an example, we observed that specific combinations of NRSE halfsite reversals may retain silencing activity. As shown in Fig. 6B, immediately upstream of the 5Ј-half motif in SCN8A is a conserved 3Ј-half motif in the reverse orientation, comparable with the partially repressing "5-3rev" shown in Fig. 4. Finally, as with classic NRSEs, some genes contain multiple clustered bNRSEs. For example, Unc5a, a netrin receptor related to axon guidance, contains two classic NRSEs in tandem in the first intron (hg17; chr5: 176,197,197,458), whereas PGSF1 (pituitary gland specific factor 1) contains two bNRSEs (hg17; chr19: 4,720,637-4,720,724), both with different 8-base spacers. Moreover, the latter overlaps an open reading frame and is marked by conservation at codon wobble positions only over the 5Ј-and 3Ј-motifs, suggesting evolutionary pressure beyond that simply for the coding sequence.

DISCUSSION
In this study, we report that a variant of the binding site for the transcription factor REST/NRSF governs expression of TPH2. The gene for TPH2 is on a different chromosome from TPH1, and their promoters lack sequence homology. Hence transcription regulation of TPH2, which is responsible for raphe serotonin biosynthesis, is likely to be important for limbic function and distinct from TPH1, an intermediate in pineal melatonin synthesis. We show that a variant of the binding element for REST/NRSF is highly conserved in the TPH2 promoter and confers transcription repression by REST/ NRSF. Our data indicate that REST binds a bipartite motif and FIGURE 5. Other genes containing bNRSEs. A, the number of bNRSEs with different spacer sizes conserved between human and mouse genomes. Representative genes with classic NRSEs (SCN2a and SCG10) and bNRSEs (all others) are labeled with respect to spacer length. B, ChIP assay for classic (BDNF and SCG10) and bNRSE binding in C6-glioma cells. Antibodies directed against REST (␣-REST) or anti-FLAG control (Control). Input material is sheared, and un-cross-linked DNA is diluted 1:100. TPH2 exon 11 is the negative control amplicon located 106 kb 3Ј to TPH2 promoter. The spacer size and relative location for each NRSE are shown at right. F, flanking region. C, results of quantitative reverse transcription-PCR for two bNRSE-containing genes in C6-glioma cells treated with Dn-REST (see "Experimental Procedures" for details). that the canonical 21-bp NRSE is only one of several potential permutations of REST binding sites. These findings suggest that TPH2 expression is regulated under the developmental program orchestrated by REST, raising the possibility that combinatorial variations of bipartite NRSE could confer differential regulation by REST.
Regulation of raphe serotonin biosynthesis is central to many models of psychopathology. Although the major focus in psychiatry has been on the anatomical distribution and signaling of serotonin receptors, it is curious that the most effective psychotropics globally increase serotonin through reuptake inhibition. Raphe tryptophan hydroxylase activity is labile and rate-limiting in serotonin biosynthesis (47). Regulation of TPH2 expression may be critical in brain serotonin availability and, consequently, vulnerability to emotional dysregulation. Recently, there has been considerable interest in TPH2 promoter polymorphisms that may be associated with psychiatric disorders (e.g. Ref. 48), and one functional polymorphism is proposed to be regulated through POU3F2 (19). Otherwise, little is known about TPH2 transcription regulation.
A number of findings indicate that the bNRSEs in TPH2 and in the other genes reported here are functional through REST. Repression of heterologous and native promoter by TPH2 bNRSE is comparable with that seen with classic SCN2a NRSE. Mutation of the conserved "GG" dinucleotide in the 3Ј-half motif derepresses T2N27, as reported for the same relative mutations in SCN21 (21). Repression of classic and bNRSEs was partially reversed by the histone deacetylase inhibitors trichostatin A and sodium butyrate. This is consistent with most studies of NRSEs in heterologous reporters, although the role of histone deacetylases at endogenous NRSEs is less clear (49,50). Bipartite NRSE failed to confer repression in mouse neuro-2A cells, a line lacking full-length REST (34), but repression was recovered with exogenous REST. As with classic NRSEs, bNRSEs in genes with predominantly neuronal expression (per expressed sequence tag expression profiles in Unigene), including CNGA3, KCNH6, BAI3, CASKIN, and CPLX1 (complexin 1), were confirmed by anti-REST chromatin immunoprecipitation in C6-glioma cells. Finally, we demonstrated by quantitative reverse transcriptase-PCR that two of the ChIP-confirmed genes, TPH2 and KCNH6, are up-regulated by Dn-REST in these cells.
Computational analyses revealed additional bNRSE targets in the genome. The flexibility of a bipartite structure predictably increased the number of potential matches by comparison with classic NRSE. However, 95% were filtered by restricting the search to human-mouse conserved motifs. We conducted the transcription regulation studies in C6-glioma cells, a rat line previously demonstrated to express full-length functional REST (34). However, we did not include the rat genome in the conservation analysis, as it is currently far less complete and inadequately annotated for genome-wide analysis. As opposed to the previously reported 1,892/1,894 consensus classic NRSEs in either genome (43), the number of classic NRSEs conserved between mouse and human genomes is only 258. The modal conserved bNRSE is of the classic length (spacer ϭ 2 bp), but there are a significant number of additional conserved bNRSEs, with a suggestive secondary peak at spacer ϭ 8 bp. Our ChIP data suggest that bNRSEs in the range of spacer ϭ 7-10 may also be functional. Combined, the bNRSEs in this range constitute an additional 147 potential targets, i.e. more than one-third of functional NRSEs may have a bipartite structure. Our search missed some NRSEs because of misalignments in the UCSC data base (e.g. M4 muscarinic cholinergic receptor) or limitations of the consensus motif (e.g. synapsin I), and we cannot rule out species-specific NRSEs; however, based on our unbiased approach, it is likely that a substantial fraction of REST/NRSF regulated genes contain a bNRSE. Interestingly, a genome-wide ChIP analysis published during revisions of this manuscript revealed similar findings (51). Six of the eight bNRSE genes that we tested by ChIP were identified in the genome-wide ChIP. However, we report a smaller number of NRSEs (both classic and bipartite) because of the additional phylogenetic conservation filter. Arguably, conserved NRSE motifs may represent more functionally relevant targets.
More studies are needed to establish the details of bNRSE binding by REST, but the current findings force a broader view of REST/NRSF function. At a minimum, it suggests that the REST/NRSF regulated transcriptome is larger and perhaps different from what was thought previously. Possibly, bNRSEs may mediate differential regulation of a third class of genes beyond the two previously suggested for classic NRSEs (27). As suggested by the examples given in Fig. 6, spacer size and halfsite inversions in bNRSE could contribute to additional complexity or preserve NRSE function in the face of evolutionary pressures. Evolutionary conservation of the spacer sequence in some genes (Fig. 6) may signify additional binding factors that facilitate or compete with bNRSE function distinct from classic NRSE. Finally, different regional and temporal binding partners could interact with 5Ј-half or 3Ј-half motifs to more finely tune silencing complex formation and neuronal cell-type specification.
The presence of a bipartite NRSE in the promoter of TPH2 is particularly intriguing because one other gene expressed in raphe cells and related to serotonergic signaling, the serotonin-1a receptor, is also thought to be regulated by REST (52). Moreover, considering the limited number of evolutionarily conserved NRSEs, it is noteworthy that a third member of the serotonin signaling system, the serotonin-6 receptor, also contains a potential conserved NRSE (supplemental Table S2). Curiously, genes for both serotonin-1a receptor and serotonin-6 receptor contain classic length (21 bp) NRSEs and are expressed in multiple neuron types (53), whereas TPH2 contains a bNRSE of 27 bp, and its expression in brain is restricted to raphe neurons (2). Possibly REST may play a global role in maturation of the serotonergic system, but the type of NRSE it binds may confer additional cell-type specific gene expression.
REST is thought to be important in dynamic (34,37) and/or long term epigenetic regulation of gene expression through recruitment of a number of chromatin remodeling factors in the vicinity of the NRSE (54). Recent studies that suggest one of these remodeling mechanisms, CpG island methylation, may be involved in maternal programming of stress-related gene expression during development (55). It is tempting to speculate that developmental events could interact with REST activity to establish the ontogenetic level of TPH2 expression and hence the vulnerability to psychopathology. The additional combina-torial possibilities afforded by a bipartite NRSE could govern subtle differences in the action of REST.