The Human Oxytocin Gene Promoter Is Regulated by Estrogens*

Gonadal steroids affect brain function primarily by altering the expression of specific genes, yet the specific mechanisms by which neuronal target genes undergo such regulation are unknown. Recent evidence suggests that the expression of the neuropeptide gene for oxytocin (OT) is modulated by estrogens. We therefore examined the possibility that this regulation occurred via a direct interaction of the estrogen-receptor complex with cis-acting elements flanking the OT gene. DNA-mediated gene transfer experiments were performed using Neuro-2a neuroblastoma cells and chimeric plasmids containing portions of the human OT gene 5'-glanking region linked to the chloramphenicol acetyltransferase gene. We identified a 19-base pair region located at -164 to -146 upstream of the transcription start site which is capable of conferring estrogen responsiveness to the homologous as well as to a heterologous promoter. The hormonal response is strictly dependent on the presence of intracellular estrogen receptors, since estrogen induced stimulation occurred only in Neuro-2a cells co-transfected with an expression vector for the human estrogen receptor. The identified region contains a novel imperfect palindrome (GGTGACCTTGACC) with sequence similarity to other estrogen response elements (EREs). To define cis-acting elements that function in synergism with the ERE, sequences 3' to the ERE were deleted, including the CCAAT box, two additional motifs corresponding to the right half of the ERE palindrome (TGACC), as well as a CTGCTAA heptamer similar to the "elegans box" found in Caenorhabditis elegans. Interestingly, optimal function of the identified ERE was fully independent of these elements and only required a short promoter region (-49 to +36). Our studies define a molecular mechanism by which estrogens can directly modulate OT gene expression. However, only a subset of OT neurons are capable of binding estrogens, therefore, direct action of estrogens on the OT gene may be restricted to a subpopulation of OT neurons.

as well as a CTGCTAA heptamer similar to the "elegans box" found in Caenorhabditis elegans. Interestingly, optimal function of the identified ERE was fully independent of these elements and only required a short promoter region (-49 to +36). Our studies define a molecular mechanism by which estrogens can directly modulate OT gene expression. However, only a subset of OT neurons are capable of binding estrogens, therefore, direct action of estrogens on the OT gene may be restricted to a subpopulation of OT neurons.
Steroid hormones are among the most potent and specific extrinsic chemical signals affecting brain function. In particular, actions of estrogens on nerve cells have been extensively studied and were shown to induce changes at the morphological, chemical and electrophysiological levels (1 these changes are thought to be mediated by regulation of specific neuronal genes. Estrogens, like other steroid hormones, bind to their cognate intracellular receptors to form steroid-receptor complexes. These complexes modulate gene transcription by interacting with specific DNA regions, referred to as estrogen-response elements (EREs)' (2-10). Gene transfer studies have identified functional EREs in the upstream regulatory regions of several estrogen responsive genes including the Xenopus and chicken vitellogenin genes (2, 4); the chicken ovalbumin (5) and apoVLDL II (6) genes; the rat prolactin (7-8) and /?-luteotropin genes (9); and the human pS2 gene (10). In spite of the pronounced effects of estrogens in the central nervous system, no EREs have so far been identified in any neuronally expressed genes. This report shows the identification and characterization of an ERE that regulates the human gene encoding the hypothalamic nonapeptide oxytocin, a neuropeptide that is largely involved in regulation of reproductive functions. Oxytocin (OT) acts as both a circulating hormone and neurotransmitter.
Circulating OT is implicated in uterine contraction and milk ejection during labor and lactation (ll-12), whereas central actions include induction of specific sexual and maternal behaviors (13). Ovarian steroids have long been known to have marked effects on OT physiology. Estrogens, in particular, stimulate OT release into the circulation (14-15) and affect morphology and electrophysiology of oxytocinergic neurons (16)(17)(18)(19). Recent studies provided evidence that estrogens are also capable of modulating OT mRNA accumulation (20-21). The precise mechanisms mediating these changes so far remained unknown. In the present study, we investigated the possibility that estrogens exert their effect via direct action on the OT gene promoter.

RESULTS AND DISCUSSION
In order to determine whether the estrogen-receptor complex could directly influence OT gene promoter activity, 381 bp of the 5'-flanking region of the human OT gene (23) were linked to the bacterial reporter gene, CAT (Ref. 22). Since there are no OT-producing cell lines, the resulting construct (POT-381CAT, Fig. lA) was transfected into the neuroblastoma derived cell line . Although this cell line does not express the endogenous OT gene, the transfected promoter elements displayed a readily detectable base-line activity in the absence of external stimulation (Fig. 1B). When these cells were endowed with estrogen receptors by co-transfection with an expression vector containing the estrogen receptor cDNA (HEO, Ref. 26), an addition of lo'-' M 17/3-estradiol elicited a 12-fold increase in expression of the chimeric gene POT-381CAT (Fig. 1B). This effect was specific since the expression of a co-transfected plasmid containing the Lac Z gene, placed under control of the SV40 early promoter, remained unaffected (Fig. 1C). Moreover, this effect was estrogen receptor-dependent, since estrogen induction could only be observed in host cells co-transfected with the estrogen receptor expression plasmid HE0 (Fig. 1).
We next characterized the dose dependence of this hormonal induction.
A dose-response relationship was established with respect to both the dose of added hormone and the dose of co-transfected estrogen receptor plasmid (Fig. 2). The minimal estradiol concentration capable of eliciting a statistically significant response was lo-' M ( Fig. 2A). On the other hand, the minimal amount of co-transfected estrogen receptor plasmid necessary to produce a significant transcriptional effect with lo-' M estradiol was as low as 200 ng (Fig.  2B). The low doses capable of induction attested to the specificity of the observed response. As a further test for specificity, we assessed the response to glucocorticoids in cells co-transfected with both estrogen and glucocorticoid receptor expression plasmids. The results depicted in Fig. 3 indicate that the response is specific for estrogens and does not occur after dexamethasone addition, although transfected cells exhibited a high dexamethasone binding capacity (334 f 29 fmol/mg protein). Moreover, addition of both steroids to cells co-transfected with both steroid receptor plasmids induced a response that did not differ significantly from the response induced by estrogens alone.
In order to test whether the observed increase in CAT activity was indeed the result of trans-activation of transcrip- tion, Sl analysis was performed. This analysis indicated, furthermore, the position at which initiation of CAT mRNA transcription occurred. As shown in Fig. 4 (lanes 1 and 2), estradiol treatment of Neuro-2a cells co-transfected with POT-381CAT and the expression plasmid HE0 resulted in a marked rise in correctly initiated CAT mRNA. Densitometric scanning of the autoradiogram indicated a 14-fold increase in band intensity.
To determine whether the 5'-flanking region of the OT gene could also confer estrogen responsiveness to a heterologous promoter in an orientation independent manner, a restriction fragment, encompassing bases -381 to -49 of the OT gene, was inserted in front of the herpes simplex thymidine kinase (tk) promoter linked to the CAT gene (ptkCAT, Ref. 24). In both orientations, the inserted fragment was able to induce a 7-fold increase in the transcriptional activity of the promoter. By contrast, ptkCAT, a plasmid devoid of any OT sequences, was unaffected by the estradiol treatment (Fig.  5).
In order to delineate more precisely the area necessary for estrogen responsiveness, 5' deletion mutants were constructed. Deletions extending beyond nucleotide -164 resulted in a significant decrease in the estrogen response (Fig.  6). Inspection of the sequence at this breakpoint revealed the presence of the sequence element 5'-GGTGACCTTGACC-3', an imperfect palindrome closely resembling the palindromic ERE consensus sequence 5'-GGTCANNNTGACC-3' (31) (difference with consensus sequence is underlined).
The present palindrome differs from the consensus ERE palindrome at position 4, where the C is replaced by a G. Previous studies (32-34) as well as our own data (Fig. 6)  CAT activity present in extracts of control cells (stippled bars) or cells treated with 10m7 M estradiol (filled bars) was determined as described under "Materials and Methods." Each bar represents the mean + S.E. of six transfections. constructs palindrome can abolish or significantly reduce estrogen inducibility.
Position 4, however, has not yet been subjected to mutational analysis. Furthermore, the naturally occurring functional EREs identified so far all contain a C in position 4. A notable exception is the first of the two ERE palindromes in the Xenopus vitellogenin gene Bl which contains a T at this position (and an A in position 1). Yet this latter palindrome is not functional on its own (35). Data from methylation interference analysis suggest that position 4 in the palindrome is indeed of functional importance, since the G which base-pairs with the C in this position forms a contact site with the estrogen receptor molecule (35). Thus, the present ERE represents a novel version of a fully functional ERE which differs at a relevant position from the consensus palindrome. It is also worthy of note that the rut OT gene 5'flanking region contains an identical sequence element at a similar position (36)(37). By contrast, the corresponding sequence element in the bovine OT gene flanking region is CATAACCTTGACC (differences underlined) (38). It would -be highly interesting to determine whether this mutated element is involved in estrogen regulation or whether another sequence element fulfills this role in the cow.
Inspection of the sequence of the human OT 5'-flanking region reveals the presence of two additional sequence elements that correspond to the right half of the ERE palindrome (TGACC). These elements are present at positions -103 and -83 (indicated by black vertical bars in Fig. 6) and have been conserved in the rat and bovine OT gene promoters where they are present at similar positions (36-38). Interestingly, an identical element, but in the opposite orientation (GGTCA), functions as an ERE in the chicken ovalbumin gene promoter (5). Either of the two constructs that contained the two TGACC elements, but not the palindromic ERE, exhibited a small but consistent l&fold response to estrogen these responses were statistically significant (p C 0.01, n = 3, for either construct), they were weak compared to the 12-fold increase mediated by the palindromic ERE at -164. Moreover, the experiments discussed in the following paragraph do not lend any support to the concept that these elements may act in synergism with the functional ERE at -164. At present, a functional significance of these elements cannot be fully excluded, since, as demonstrated by Tora et al. (5), the function of this half palindrome may depend on the presence of additional cell-specific transcription factors. A definite answer has thus to await the availability of a cell line expressing the endogenous OT gene.
For several steroid inducible genes, a synergistic action between steroid receptors and other trans-acting factors has been demonstrated (7,(39)(40)(41)(42). Therefore, we asked whether neighboring sequences on the promoter might serve as binding sites for synergistically acting transcription factors. Since removal of 217 bases upstream of the ERE palindrome did not affect estrogen inducibility (Fig. 6), we focused next on sequences downstream of this element. In addition to the two half-palindromes mentioned previously, this promoter area includes the following sequence elements which display considerable sequence conservation through species evolution: (i) a TATA box equivalent (CATAAAA) beginning at -28; (ii) a CCAAT box at -80; and (iii) the motif CTGCTAA at -139 that is reminiscent of the heptamer CTGATAA. This latter element, termed "elegans box," is thought to play a supportive role in estrogen control (2) and is closely associated with functional EREs in the vitellogenin genes of several species including the nematode Caenorhabditis &guns (43). However, either removal of the elegans box alone (pOTAebCAT) or replacement of sequences -145 to -50 by unrelated spacer sequences (POT-49+eCAT) did not significantly affect estrogen induced expression of the CAT reporter gene (Fig. 7). In the latter construct, the elegans box, as well as the CCAAT box and the two ERE half-palindromes, were all replaced by vector sequences. Sl analysis revealed that even in the truncated 5'-flanking region of POT-49+eCAT, initiation oc-  (39). The latter suggests functional cooperativity between the CCAAT box binding factor and the steroid-receptor complex. In the present study, the ERE, located 164 bp from the transcription start site, is fully active in the absence of a CCAAT motif and in the absence of the two ERE half-palindromes that could conceivably substitute for an additional ERE (POT-49+eCAZ', Fig. 7). The only sequence requirements for full inducibility are present in a short promoter fragment that contains the TATA box (-49 to +36). It remains to be determined whether this small region contains other specific sequence elements that act in synergism with the ERE. By contrast, any cooperative action of the elements located between -145 and -50 may require cell specific tram-acting factors that are absent from Neuro-2a cells.
In conclusion, the present data defines a molecular mechanism by which estrogens can directly modulate OT gene expression.
The existence of such a mechanism has been suggested by a series of in uiuo observations which established a relationship between estrogen levels, on the one hand, and neuronal OT mRNA or peptide accumulation on the other (20-21, 44-47). This concept has been supported further by the demonstration that oxytocinergic neurons are capable of concentrating exogenously added labeled estrogens (48-49). However, since the ability to concentrate estrogens is apparently limited to only a subset of oxytocinergic neurons (4% 49), direct action of estrogens on the OT gene may be restricted to a subpopulation of OT neurons. Continued investigation at the molecular level of steroid regulation of neuronal gene expression should shed further light on the complex actions of steroids on brain function.