help button home button Endocrine Society Endocrine Reviews
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS

This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Purchase Article
Right arrow View Shopping Cart
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow Request Copyright Permission
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Andersen, B.
Right arrow Articles by Rosenfeld, M. G.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Andersen, B.
Right arrow Articles by Rosenfeld, M. G.
Endocrine Reviews 22 (1): 2-35
Copyright © 2001 by The Endocrine Society

POU Domain Factors in the Neuroendocrine System: Lessons from Developmental Biology Provide Insights into Human Disease1

Bogi Andersen and Michael G. Rosenfeld2

Division of Endocrinology and Metabolism (B.A.), and Howard Hughes Medical Institute (M.G.R.), Department of Medicine, University of California, San Diego, La Jolla, California 92093-0648


    Abstract
 Top
 Abstract
 I. Introduction
 II. Structure and...
 III. Development of the...
 IV. Expression and Function...
 V. Relevance of POU...
 References
 
POU domain factors are transcriptional regulators characterized by a highly conserved DNA-binding domain referred to as the POU domain. The structure of the POU domain has been solved, facilitating the understanding of how these proteins bind to DNA and regulate transcription via complex protein-protein interactions. Several members of the POU domain family have been implicated in the control of development and function of the neuroendocrine system. Such roles have been most clearly established for Pit-1, which is required for formation of somatotropes, lactotropes, and thyrotropes in the anterior pituitary gland, and for Brn-2, which is critical for formation of magnocellular and parvocellular neurons in the paraventricular and supraoptic nuclei of the hypothalamus. While genetic evidence is lacking, molecular biology experiments have implicated several other POU factors in the regulation of gene expression in the hypothalamus and pituitary gland. Pit-1 mutations in humans cause combined pituitary hormone deficiency similar to that found in mice deleted for the Pit-1 gene, providing a striking example of how basic developmental biology studies have provided important insights into human disease.

I. Introduction

II. Structure and Transcriptional Function of POU Domain Factors

A. DNA-binding sites for POU domain factors

B. Protein chemistry

C. Protein-protein interactions

D. Mechanisms of transactivation

III. Development of the Hypothalamo-Pituitary Region

IV. Expression and Function of POU Domain Factors in the Neuroendocrine System

A. Pit-1

B. Oct-1

C. Oct-2

D. Overview of expression of Brn-1, Brn-2, Brn-4, and Tst-1 in the neuroendocrine system

E. Brn-2

F. Brn-4

G. Tst-1

H. Oct-3

I. Brn-5 and RPF-1

V. Relevance of POU Domain Factors to Diseases of the Neuroendocrine System

A. Pit-1 mutations as a cause of combined pituitary hormone deficiency in humans

B. Brn-4 mutations in humans

C. Potential implications for other diseases.


    I. Introduction
 Top
 Abstract
 I. Introduction
 II. Structure and...
 III. Development of the...
 IV. Expression and Function...
 V. Relevance of POU...
 References
 
THE ACRONYM POU (pronounced pow) is derived from the names of three mammalian transcription factors, the pituitary-specific Pit-1 (1, 2), the octamer-binding proteins Oct-1 (3) and Oct-2 (4, 5, 6, 7), and the neural Unc-86 from Caenorhabditis elegans (8). These initial members of the POU gene family, as well as subsequently discovered members, are characterized by a unique bipartite DNA-binding domain referred to as the POU domain (9). This domain is composed of a variant homeodomain (POU-H), which is linked via a nonconserved linker to another conserved domain referred to as the POU-specific (POU-S) domain. Both subdomains contain helix-turn-helix motifs that directly associate with the two components of bipartite DNA-binding sites (Fig. 1Go).



View larger version (55K):
[in this window]
[in a new window]
 
Figure 1. The three-dimensional structure of the Pit-1 POU domain bound to DNA compared with the Oct-1 POU domain bound to an octamer site. Pit-1 binds as a dimer to a palindromic PRL-related DNA site with the POU-S and POU-H domains of each monomer binding to the same side of DNA (left panel). The third helix of both subdomains makes major groove contacts. For comparison, the schematic on the right represents the different organization of Oct-1 bound to an octamer site where the two subdomains bind on the opposite sites of DNA and the Oct-1 POU-S domain (shown in yellow) has a flipped orientation compared with the Pit-1 POU-S domain (shown in pink).

 
While highly homologous in the POU domain, the genes expressing the various POU domain factors are distributed on several different chromosomes, not clustered like the Hox genes (10, 11, 12, 13, 14, 15, 16, 17, 18, 19). POU domain genes have been described in organisms as divergent as C. elegans (20), Drosophila (21, 22, 23), Xenopus (24), zebrafish (25, 26), and humans, but have not yet been identified in plants or fungi. In all species these genes seem to carry out essential functions for organ development and cellular differentiation.

Although some general conclusions can be reached about POU domain factors, more importantly the various members serve a wide variety of functions during development. In fact, it is reasonable to predict that an individual POU factor may play multiple distinct roles in those organs in which it is expressed during development and in the adult. There is evidence for the role of POU domain factors in inhibition and promotion of cell proliferation and determination of cell lineages, as well as in regulation of cell migration, survival, and terminal differentiation. Because of their overlapping expression patterns and similar DNA-binding preferences, POU factors coexpressed in a single organ may be functionally redundant. In addition to a role in early and late development, members of this family play roles in physiological regulation of gene expression in the adult. While POU domain genes are expressed in all organ systems, this review focuses on the role of POU domain factors in the neuroendocrine system with special emphasis on the hypothalamo-pituitary axis.

A great deal of knowledge has accumulated about the biological function of POU domain factors since their initial identification in 1988. In addition to insightful in vitro experiments, progress in this field has been greatly facilitated by the ability to create targeted mutations of POU domain genes in the mouse. These studies, as well as the discovery of human disease mutations involving the POU domain factor Pit-1, have established a crucial role for POU domain factors in neuroendocrine development and function.

The 15 known mammalian POU domain genes have been classified into six different groups based on homology in the POU domain region, including the nonconserved linker region (Table 1Go). The class I POU factor, Pit-1 (GHF-1), is required for the generation and maintenance of three different cell types in the anterior pituitary gland of humans and other mammals (1, 2, 27). Expression analyses and in vitro biochemical/molecular biology experiments have implicated the octamer-binding proteins Oct-1 and Oct-2, which belong to class II POU domain factors, in neuroendocrine function and development (28, 29, 30, 31). The third class II factor, Skn-1a, is selectively expressed in epidermis and has no role in the neuroendocrine system (32, 33, 34, 35). The class III POU domain subgroup consists of four closely related factors, Brn-1, Brn-2, Brn-4, and Tst-1/Oct-6/SCIP, which are all encoded by intronless genes (36). For members of this subgroup, a role in neuroendocrine development has been most clearly established for Brn-2, which is required for the formation of hypothalamic nuclei and the posterior pituitary gland (37, 38). The other class III POU factors may also play a crucial role in neuroendocrine function and development. Brn-3 and two other members of the class IV POU factors have been shown to be important for various aspects of sensory development, but have no known roles in the neuroendocrine system and will not be further discussed in this review (39, 40, 41, 42, 43, 44, 45, 46, 47, 48). The class V POU factor Oct-3 (Oct-4) is expressed in germ cells, embryonic stem cells, and in early neurogenesis where its down-regulation may be crucial for proper neural formation (49, 50, 51, 52, 53). Sprm-1, the second class V factor, is selectively expressed in spermatocytes and has no role in the neuroendocrine system (54, 55). The class VI POU factor Brn-5 is widely distributed and highly expressed in hypothalamic regions (56, 57, 58). The other class VI member, retina-derived POU-domain factor-1 (RPF-1), is expressed in the retina, the medial habenula, and in dispersed populations of neurons in the dorsal hypothalamus (59).


View this table:
[in this window]
[in a new window]
 
Table 1. Mammalian POU domain factors

 
In addition to reviewing the demonstrated or proposed roles for POU domain factors in the neuroendocrine system, we will summarize aspects of the basic molecular biology of POU factors. Extensive investigations into the biochemistry and molecular biology of POU domain factors have given insights into structure-function aspects such as DNA binding, protein-protein interactions, and mechanisms of transcriptional regulation. While many of these studies involve the action of POU domain factors outside the neuroendocrine system, they are informative and likely to provide leads regarding the mechanisms of action for POU domain factors in the neuroendocrine system.


    II. Structure and Transcriptional Function of POU Domain Factors
 Top
 Abstract
 I. Introduction
 II. Structure and...
 III. Development of the...
 IV. Expression and Function...
 V. Relevance of POU...
 References
 
A. DNA-binding sites for POU domain factors
While POU domain factors exhibit a surprising flexibility in DNA binding (60, 61), a general rule has emerged suggesting that high-affinity binding sites are often bipartite with each "half-site" contacting either the POU-S or the POU-H domain (Fig. 1Go) (62, 63). This principle has been elegantly demonstrated for the interaction of Oct-1 with an octamer site (ATGCAAAT) where the POU-S domain and the POU-H domains associate with the ATGC- and AAAT-parts, respectively (64, 65, 66). The spacing and orientation between the two halves of a binding element can vary, and the ability of different POU domain factors to recognize such sites depends in part on the length, sequence, and flexibility of the linker tethering the POU-S and POU-H domains (63, 67). In addition, there is evidence that diversity in the POU DNA binding may stem, in part, from amino acid-base interactions that are more flexible than those observed for classical homeodomains (62).

Often, high-affinity binding sites for POU factors are flanked by lower affinity binding sites that are capable of promoting cooperative DNA binding of POU factor dimers. These "second" sites are usually highly diverged from high-affinity binding sites and may contain isolated A/T-rich regions or the POU-S recognition sequence ATG (CAT on the opposite strand). In this section, we will first review the DNA binding characteristics of different POU domain factors, and then summarize the main conclusions from the collection of studies on this issue.

1. DNA binding sites for the class I factor Pit-1. Pit-1 was originally identified based on its ability to bind to multiple sites in the regulatory regions of its target genes GH and PRL (1, 2). Similar Pit-1 binding sites have been described in the regulatory regions of several other pituitary-expressed genes that are likely targets of Pit-1. These include the Pit-1 gene itself (68, 69, 70), the TSH{beta} gene (71, 72, 73), the rat and human GHRH receptor gene-regulatory regions (74, 75), the rat somatostatin receptor 1 gene-regulatory region (76, 77), and the human TRH receptor gene-regulatory region (78). These sites, which vary widely around the consensus sequence ATGNATA(A/T)(A/T)(A/T), tend to be found in multiples in Pit-1 target genes (Table 2Go). On most sites, Pit-1 appears to bind as a homodimer with one Pit-1 molecule contacting the high-affinity octamer-related sequence, ATGNATA, and the second molecule binding cooperatively to the 3' located lower affinity A/T-rich region (79, 80, 81). On some sites Pit-1 can bind as a heterodimer with the POU domain factor Oct-1, and on yet other sites it binds as a monomer (82).


View this table:
[in this window]
[in a new window]
 
Table 2. Pit-1 DNA-binding sites

 
While Pit-1 also binds to the classical octamer site, it does so with lower affinity than to natural sites, and at least in one experiment seemed incapable of transactivating through octamer sites (79).

2. DNA binding sites for the class II factors Oct-1 and Oct-2. The octamer DNA element, ATGCAAAT, frequently found within gene-regulatory regions, is conserved and functionally important for transcriptional regulation of several genes. These include the B cell-specific immunoglobin genes (83), the ubiquitously expressed small nuclear (sn)RNA genes (84, 85), the cell cycle-specific histone H2B genes (86), and a host of other widely and selectively expressed genes (Table 3Go). Early studies attempting to identify the transcription factors binding to octamer elements led to the isolation of Oct-1 and Oct-2 as the major components of octamer-binding activity in eukaryotic cells. Review of the literature indicates that the Oct-1 POU domain has a strong preference for the core element, ATGCAAAT, but clearly several variants of the core element are compatible with high-affinity DNA binding (Table 3Go). Experiments have also indicated that the bases flanking the core octamer provide additional binding specificity (60).


View this table:
[in this window]
[in a new window]
 
Table 3. Oct-1/2 DNA-binding sites

 
Two studies using random DNA-binding selection methods found that Oct-1 can bind to several different DNA sites with high affinity (61, 65), consistent with previously described broad binding specificity for Oct-1 (60). Yet, when the binding sites were aligned, both groups identified an octamer consensus sequence with position 1 (A) and 7 (A) nearly invariant in sites that bear resemblance to octamer sites (65). Both groups found that a "T" in position 1 was common and that a common octamer variant was a "T" instead of an "A" in position 5 (61, 65). Review of the literature describing Oct-1/2 binding sites in gene-regulatory regions of diverse genes also shows a high frequency of octamer-like sequences (Table 3Go). It is striking that alignment of these sites results in a consensus sequence remarkably similar to that identified in the two random site selection studies (Table 3Go). However, this approach is unlikely to give a true representation of the type of sites found in gene-regulatory regions because in their studies, investigators are much more likely to identify and report on the well known octamer site rather than more divergent sites.

Oct-1 can also bind to a distinct site, TAATGARAT (R is a purine), found in the herpes simplex virus immediate early (HSV IE) promoters (87, 88). The affinity of Oct-1 alone for this site is much lower than for that of the octamer element (64), but in contrast to the classical octamer site, this site can promote the formation of a complex of Oct-1 and VP16, which is required for viral transcription (Fig. 2Go). The critical feature of this site is the 3'-GARAT motif, which is necessary for the recruitment of VP16, possibly by a mechanism involving the recognition of specific conformations of Oct-1 (62, 89) or direct nucleotide recognition (90). On this site, the POU-S domain is located on the opposite side of the POU-H domain as compared with Oct-1 bound to a classical octamer site (62).



View larger version (31K):
[in this window]
[in a new window]
 
Figure 2. Selective protein-protein interactions involving POU domain factors. Interactions of POU domain factors with other proteins are influenced both by the amino acid sequence of the POU domain and the sequence and context of the DNA binding site. a, Pit-1 interacts with GATA transcription factors. On promoters containing binding sites for both Pit-1 and GATA, the interaction seems to be cooperative and results in transcriptional activation (left panel). On promoters containing GATA-binding sites, but lacking Pit-1 binding sites, Pit-1 can inhibit GATA binding to DNA and transactivation. b, The cofactors VP16 and OCA-B interact selectively with Oct-1 and Oct-2. VP16 interacts with Oct-1 on TATGARAT DNA sites, with VP16 contacting the GARAT part of the site. This interaction is stabilized by HCF. Oct-1 and VP16 do not interact on a classical octamer site and VP16 is incapable of interacting with the highly related Oct-2, even on TATGARAT sites, indicating that minor differences in the sequence of the POU domain can direct the specificity of protein-protein interactions. In contrast, OCA-B can interact both with Oct-1 and Oct-2. However, this interaction is restricted to sites related to the classical octamer sequence and does not form on TATGARAT sites.

 
In addition, Oct-1 can bind to the heptamer element CTCATGA found next to the octamer site in immunoglobin heavy chain promoters (91, 92, 93). While divergent from the octamer site, both Oct-1 and Oct-2 can recognize this site, albeit with lower affinity. However, when Oct-1 or Oct-2 occupies the adjacent octamer site, binding to the heptamer site is strongly enhanced (91, 92, 93). Cooperative binding to the heptamer element occurs when the spacing between the two elements is either 2 or 14 bp, suggesting that the coordinated binding is caused by protein-protein interactions (91). This cooperative binding to the heptamer site contributes to transactivation by Oct-1 and Oct-2 (91, 93).

Several divergent sites capable of binding Oct-1 have been identified in diverse genes, including the GnRH gene-regulatory region (28, 29) (Table 3Go). An unusual binding site has been described in the human interleukin 5 (IL-5) promoter. In this site, inverted POU-S binding sites seem to flank two POU-H domain binding sites (94, 95). Yet other sites are similar to the TAATGARAT site in that they only contain recognizable POU-H domain binding sites (Table 3Go). The binding specificity of Oct-1 and Oct-2 has been found to be indistinguishable in several studies, and other studies indicate that the third class II POU factor, Skn-1a, has identical binding preferences (35, 96). These observations suggest that within a POU class, binding preferences tend to be similar.

3. DNA binding sites for class III factors. Studies have shown that the class III POU factor Tst-1 binds to and activates octamer site-dependent promoters (97, 98). Similar to Oct-1, different binding affinities were observed with perfect octamer sites containing distinct flanking sequences, indicating that the flanking sequences are also important determinants of binding affinity for Tst-1 (98). Tst-1 also binds to five distinct sites in the myelin-specific P0 promoter with a consensus sequence, GA(A/T)T(T/A)ANA, which appears unrelated to the octamer site (99). The rat GnRH promoter contains three Tst-1 binding sites that appear to conform to the consensus derived from the Tst-1 binding sites in the P0 promoter (100) (Table 4Go). Tst-1 also binds with high affinity to two sites in the JC virus promoter, but only one of the sites confers transcriptional activation. While they are AT-rich, neither site bears any similarity to octamer sites nor the P0 consensus site (101).


View this table:
[in this window]
[in a new window]
 
Table 4. Class III DNA-binding sites

 
Brn-2, like Tst-1, can bind to and activate promoters dependent on octamer sites (102). Rhee et al. (103) employed a random site selection method to determine optimal binding sites for class III POU factors, using Brn-2 as a prototype member. Their study confirmed the ability of class III factors to interact with octamer-like sites, especially those that have an associated heptamer site. In addition, Brn-2 was found to have preference for sites with the consensus ATG(A/C)AT(A/T)0–2ATTNAT, and on these sites, Brn-2 bound as highly cooperative dimers (103). Brn-2 binds to a divergent site in the neuronal promoter of the aromatic L-amino acid decarboxylase gene, a sequence that overlaps with a binding site for winged helix/forkhead protein HNF-3 (104).

Brn-4 DNA binding preferences have also been evaluated by a random site selection method where it was found that the optimal binding site, CAATATGCTAAT, is related to the smaller octamer site, containing a TAAT instead of the AAAT (105). Thus, the optimal binding sites selected with Brn-2 and Brn-4 are different. It is quite possible that these differences are due to methodological variables rather than bona fide binding differences between these two class III POU factors. Since it is likely that all POU domain factors can bind to DNA in different conformations, it is not surprising that different studies have identified distinct binding sites for class III POU factors. Nonoctamer DNA-binding sites for Brn-4 have been found in the striatal D1 dopamine receptor (106) and proglucagon genes (107). These sites seem to be most similar to the Tst-1 sites in the P0 and rGnRH promoters (Table 4Go).

4. DNA binding sites for the class V factor Oct-3. Oct-3 binds with high affinity to octamer sites (108, 109, 110), but also to so called palindromic Oct factor recognition elements (PORE), which simulate a pair of inverted homeodomain sites separated by 5 bp, ATTTG +5 CAAAT, found in the osteopontin enhancer (Table 5Go). In addition to Oct-3, other Oct factors can bind to this site, either as homodimers or heterodimers (111). Oct-3 is reported to transactivate more robustly from this element than a classical octamer site (111).


View this table:
[in this window]
[in a new window]
 
Table 5. Oct-3 DNA-binding sites

 
Oct-3 can also bind to and regulate transcription through an element located in the regulatory region of a gene encoding the embryonic stem cell coactivator UTF1. This element, which is one base different from a classical octamer site (Table 5Go), selectively recruits Oct-3 and Sox-2 and is incapable of binding complexes containing Oct-1 and Tst-1 (112).

5. DNA binding sites for the class VI factor Brn-5. In experiments using a random site selection method, Brn-5 was found to prefer the binding site GCATAA(T/A)TTAT, which is related to the class IV factor Brn-3 optimal binding site (103). In different experiments where Brn-5 site selection was performed with a fixed site, GCAT, followed by eight random nucleotides, an octamer-like sequence was selected: GCATATGATAAT (56). However, it is clear that Brn-5 binds poorly to a classical octamer site (56). Alignment of three high-affinity sites suggested a high-affinity consensus site GCATN2–3TAAT (56) (Table 6Go).


View this table:
[in this window]
[in a new window]
 
Table 6. Brn-5 DNA-binding sites

 
6. Conclusions. DNA-binding sites capable of binding POU domain factors have been identified with two distinct methods. Studies aimed at determining important gene-regulatory elements have identified several DNA sites that bind POU domain factors important for regulation of the gene in question. Genetic evidence suggests that the Pit-1 binding sites in the GH, PRL, and TSH{beta} genes are bona fide response elements. For most other sites identified in such studies, genetic evidence for their importance is lacking. An alternative approach has been to use a selection of high-affinity binding sites from a set of random sites in vitro. While this approach may identify high-affinity sites, those are not necessarily biologically important sites. However, from these two approaches several general conclusions about POU domain binding sites can be reached.

First, members of the same subclass of POU domain factors tend to have similar DNA-binding preferences, most clearly demonstrated for class II POU factors. Second, while distinct POU classes may have somewhat distinct DNA-binding preferences, a single POU factor can recognize several different DNA-binding sites. Third, POU factors of different classes can often recognize the same element with similar affinity, and all POU factors discovered so far seem to be capable of binding to octamer sites with reasonably high affinity. Fourth, while the core recognition sequence for monomeric binding is often 8 bp long, bases outside the core region may be important for binding affinity and transcriptional activity. Fifth, while POU domain factors can bind as monomers, some factors, such as Pit-1 and class III factors, can homodimerize, especially on appropriate symmetrical sites. Sixth, while transcriptional activation may correlate with binding affinity on monomeric sites, other features of the site, such as the ability to allow dimerization and interaction with other factors, may be important determinants of whether a particular site confers activation (Fig. 2Go). Seventh, the DNA-binding site can determine the ability of a bound POU factor to recruit associated proteins (Fig. 2Go). Finally, target genes that are responsive to POU domain factors frequently have a cluster of POU-binding sites, particularly when the regulatory regions are located at a distance from the promoter.

B. Protein chemistry
The structure of the POU domain has been extensively studied for both Oct-1 and Pit-1, using NMR (113, 114, 115, 116) and crystallographic (66, 117) methods. The similarity in structure between these two POU domains strongly suggests that the basic observations from these studies can be applied to other POU domains. The POU domain is composed of a 74- to 82-amino acid POU-S domain and a 60-amino acid POU-H domain with a 15- to 56-amino acid flexible linker tethering the two domains (Fig. 1Go). The two subdomains are structurally independent, and individually they are capable of low-affinity DNA interactions (64). High-affinity DNA binding, however, requires the intact POU domain, with each subdomain making major groove DNA contacts. The structure of the POU-S domain is characterized by four {alpha}-helices surrounding a hydrophobic core similar to the helix-turn-helix motif of bacteriophage {lambda} and 434 repressor, and 434 Cro DNA-binding domains (113, 115). While somewhat divergent, the POU-H domain, like classic homeodomains, exhibits a helix-turn-helix structure with three {alpha}-helices (118, 119, 120). For both the POU-S and POU-H domains, the second and third helix form a helix-turn-helix structure with the third helix docking to DNA in a major groove where amino acids making direct base contacts tend to be conserved among POU domain factors (121, 122).

The two subdomains do not contact each other, and the linker is not visible in the cocrystals, suggesting that it may be unstructured, perhaps providing flexibility to the POU domain (Fig. 1Go). This flexibility would allow the two subdomains to acquire different orientation and spacing relative to each other, presumably explaining the versatility in DNA sites that can bind POU domain factors, and the variability in surfaces available for interactions with coregulatory proteins (63). This flexibility may depend, in part, on the length of the linker as evidenced by results showing that the Brn-2 POU domain can adopt two different orientations on DNA, while this was not observed with Brn-3, which contains a shorter linker (62, 63). An effect of the linker on DNA binding specificity was also found in experiments comparing binding of chimeric molecules from the Oct-1 and Pit-1 POU domains (123). It has been shown that optimal binding of Oct-1 to an octamer site requires a minimum linker length between 10 and 14 amino acids (67). Oct-1 molecules with a shorter linker bind DNA only if the POU-S recognition site is inverted in orientation, apparently bringing it closer to the POU-H domain. In addition, mutations of some conserved residues in the Oct-1 linker altered DNA binding, indicating that there are both sequence and length requirements for the function of the linker (67).

An example of the subdomain flexibility is provided by the binding of Oct-1 to the octamer element (ATGCAAAT) where the POU-S domain recognizes the 5'-ATGC-part and the POU-H domain recognizes the 3'-AAAT-part. On the TAATGARAT site from the herpes simplex virus (HSV) IE promoter, the Oct-1 POU-H domain recognizes the 5'-TAAT part, and the POU-S domain recognizes the 3'-GARAT part in a flipped orientation as compared with Oct-1 binding to an octamer site (62).

While the primary, secondary, and tertiary structures are similar, crystallographic studies of Oct-1 and Pit-1 on two different DNA sites have shown quite distinct structures as to the relationship between the two POU subdomains (66, 117) (Fig. 1Go). The Oct-1 POU domain binds to opposite faces of an octamer DNA-binding site (the histone H2B promoter octamer). In contrast, the Pit-1 subdomains bind to perpendicular faces of a palindromic DNA site (derived from the PRL 1P site), with the POU-S domain exhibiting flipped orientation compared with Oct-1 structure (Fig. 1Go). In addition, while Oct-1 binds as a monomer to an octamer site, Pit-1 forms homodimers on the palindromic PRL 1P-related site via interactions between the C terminus of the DNA-recognition helix of the homeodomain and helix {alpha}1 and the loop between helices {alpha}3 and {alpha}4 of the POU-S domain. Most likely, these differences in structure between Oct-1 and Pit-1 are due to the different DNA-binding sites used for cocrystallization, which is supported by data indicating that the Pit-1 POU domain binds as a monomer to a classical octamer site (124).

It has been shown that POU factors can be modified by phosphorylation. Oct-1 and Pit-1, for example, are phosphorylated on homologous residues in the POU domain, Ser-385 and Thr-220, respectively, and this phosphorylation leads to decreased DNA binding to octamer and GH POU sites (125, 126). On other sites, such as those found in the TSH{beta} and PRL promoters, phosphorylation by protein kinase A or C enhances DNA binding (127, 128). While cAMP-dependent kinase can phosphorylate these sites in vitro, the identities of the relevant kinases operating in vivo are unclear. For Oct-1, this phosphorylation occurs in a cell cycle-specific manner, thus explaining the S phase-specific activation of the histone H2B gene by Oct-1 (86, 125). Jun kinase (JNK)/stress- activated protein kinase can bind to the activation domain of Brn-5 and phosphorylate its POU domain on sites that are distinct from the aforementioned Oct-1/Pit-1 phosphorylation sites. This phosphorylation leads to increased binding to DNA and enhanced transactivation (129). Finally, it has been demonstrated that Pit-1 can be acetylated by CREB-binding protein (CBP), but the biological significance of this finding remains to be determined (130).

C. Protein-protein interactions
POU domain factors regulate transcription by interacting with other proteins. POU interacting proteins can be classified into four classes: DNA-binding transcriptional activators, coregulators, basal factors, and replication factors (131). A prominent feature of POU factors is their ability to form cooperative homodimers, and even heterodimers with other POU factors, especially when bound to the appropriate DNA site (80, 82, 132) (Fig. 2Go). POU factors also have a striking ability to interact with a variety of divergent structural domains in heterologous factors, including other DNA-binding proteins and coregulators that are incapable of independent DNA binding. In this chapter we will review protein-protein interactions involving those POU factors expressed in the neuroendocrine system (Table 7Go). Many of these interactions may not be directly relevant to regulation of gene expression in the hypothalamo-pituitary axis, but the principles derived from these studies are likely to apply to studies of POU domain factors in the neuroendocrine system (Fig. 2Go).


View this table:
[in this window]
[in a new window]
 
Table 7. Protein-protein interactions involving POU domain factors

 
1. Pit-1 interactions.
a. Interactions with DNA-binding proteins.
Pit-1 forms highly cooperative homodimers on the natural binding sites in regulatory regions of the GH, PRL, and TSH{beta} genes (80). This Pit-1 homodimerization has been shown in living cells with fluorescence resonance energy transfer microscopy (FRET) (133). Functional experiments indicate that the POU-S domain is important for Pit-1-Pit-1 interactions and that homodimerization is important for efficient transactivation (80). The ability of some natural Pit-1 mutants to inhibit Pit-1 action in a dominant negative fashion further supports the importance of Pit-1 homodimerization in gene activation (see Pit-1 mutations as a cause of combined pituitary hormone deficiency in humans). In addition, Pit-1 can form heterodimers with Oct-1, which results in synergistic transactivation of the PRL promoter (82).

Pit-1 can also interact to varying degree with several heterologous DNA-binding proteins which, when bound to the same promoter, often lead to synergistic transactivation. These interacting factors include the nuclear hormone receptors TR (thyroid hormone receptor), RAR (retinoic acid receptor), RXR (retinoid X receptor), ER (estrogen receptor), VDR (vitamin D receptor), and GR (glucocorticoid receptor). Pit-1, for example, interacts strongly with RXR and more weakly with TR and RAR, which may play a role in synergistic regulation of the GH promoter (134, 135, 136). Pit-1 can also interact with ER (137, 138, 139, 140) and VDR (141) in synergistic activation of the PRL gene. The human PRL promoter is inhibited by glucocorticoids, mediated by a gene regulatory region containing Pit-1 binding sites but no glucocorticoid response element (GRE). Studies show that GR and Pit-1 can interact in solution and that the inhibition does not require GR to bind to DNA (142).

Pitx1, a bicoid-like homeodomain factor, interacts with the N terminus of Pit-1 (143) and synergistically transactivates the PRL promoter with Pit-1 (144, 145). A direct interaction of Pit-1 with the related Pitx2 removes an inhibitory effect of a 39-amino acid C-terminal sequence of Pitx2, thus promoting DNA binding and transactivation (146). Pit-1 also interacts in solution with the LIM homeodomain protein P-Lim (Lhx3) and acts synergistically with P-Lim to regulate the PRL enhancer/promoter (147).

Synergistic interactions between Ets-1 and Pit-1 may be important in regulation of the PRL promoter. The synergistic gene activation requires direct Pit-1-Ets-1 protein-protein interaction and is not observed with Ets-2 (148). This interaction, which has been shown in living cells with FRET microscopy (133), maps to the POU domain (149). Pit-1{beta}, an isoform of Pit-1 containing 26 additional amino acids in the N terminus, is incapable of transcriptional synergism with Ets-1, apparently because of additional interaction surface that maps to the N-terminal sequence specific to this isoform (149).

Interactions between Pit-1 and the transcription factor GATA-2 are more complex and have distinct features depending on whether both factors bind to DNA (150, 151) (Fig. 2AGo). Pit-1 and GATA-2 can interact in solution (150, 151), and the interaction interface maps to the POU-H domain of Pit-1, specifically the N-terminal basic residues and the surface of the second helix (P26, Q29) (151). In GATA2, the C-terminal DNA-binding zinc finger and an adjacent cluster of basic residues seem to be important. Pit-1 can inhibit binding of GATA-2 to cognate DNA sites on promoters that do not have Pit-1 binding sites, such as the gonadotrope-specific SF-1 promoter (151) (Fig. 2AGo). This seems to be important for generation of the gonadotrope phenotype. In contrast, Pit-1 leads to synergistic activation with GATA-2 on promoters containing adjacent Pit-1 and GATA-2 sites, such as the thyrotrope-specific TSH{beta} promoter (150, 151) (Fig. 2AGo). As described later, these interactions may be important for the formation of both thyrotrope and gonadotrope cell types.

b. Interactions with coregulator proteins.
In addition, Pit-1 has been demonstrated to interact with coregulators. The POU domain of Pit-1 interacts with CREB binding protein (CBP) through two cysteine-histidine rich domains (C/H1 and C/H3) (152, 153) (Fig. 3Go). Pit-1 can also, via its POU-H domain, interact with the corepressor N-CoR (nuclear receptor and corepressor) (153). As will be described later, these interactions may be important for transactivation and the role of Pit-1 in signal transduction.



View larger version (21K):
[in this window]
[in a new window]
 
Figure 3. A schematic model describing mechanisms of transcriptional regulation by Pit-1. Pit-1, which binds as a dimer to DNA, can associate with the CREB-binding protein (CBP)/p300 complex mediating histone acetylation and gene activation. CBP may also mediate cAMP responses on Pit-1 target genes. Alternatively, Pit-1 may associate with complexes that mediate histone deacetylation and repression of gene expression. Presumably, differential association with these two complexes can determine the transcriptional outcome of Pit-1 binding to a promoter region. In addition, the N terminus of Pit-1 participates in transcriptional activation, perhaps by making direct contacts with the core transcriptional machinery. In addition, several factors, such as the DNA site proteins bound to adjacent DNA sites and modifications of Pit-1 itself, contribute to transcriptional regulation.

 
2. Oct-1 and Oct-2 interactions.
a. Interactions with DNA-binding proteins.
The Oct-1 POU domain forms transient homodimers in solution in vitro, requiring both the POU-S and the POU-H domains (132). This interaction, which is relatively weak, is stabilized by binding to the octamer-heptamer site, consistent with other experiments showing that this site mediates cooperative binding (132). Oct-1 can also heterodimerize with other POU factors, including Pit-1 (82), Oct-3 (154), Oct-2, and Tst-1 (132).

Both Oct-1 and Oct-2 have been shown to interact with DNA-binding factors from other classes, including nuclear hormone receptors, such as RXR (138). The interaction interfaces seem to involve the Oct-1 POU-H domain, but not the POU-S domain, and the DNA-binding domain and hinge region of RXR. Functional studies suggest that Oct-1 may negatively regulate nuclear signaling pathways by altering the DNA-binding ability of nuclear receptor-RXR heterodimers. In contrast, on artificial promoters containing vitamin D response elements and octamer sites, Oct-1 and the vitamin D receptor exhibit cooperative binding and transcriptional synergism mediated by the DNA-binding domain of the vitamin D receptor (155).

In solution, the POU domains of Oct-1 and Oct-2 can bind to the DNA-binding domain of GR, including an interface that contains amino acids C500 and L501 (156). Again the POU-H domain of Oct-1 seems to be responsible for the interaction (156). Microinjection experiments in zebrafish embryos suggest that such DNA-independent effects of GR involving interactions with the Oct-1 and Oct-2 POU-H domains are biologically important (157). The GR-Oct-1 interaction can both modulate the transcriptional activity of Oct-1 on certain promoters, like the GnRH promoter as detailed later (158), and the activity of GR on its target genes (138). On the mouse mammary tumor virus (MMTV) promoter, it has been shown that Oct-1 and GR bind in a cooperative fashion and that this leads to a synergistic transcriptional regulation on naked DNA templates in vitro (159). Oct-2 can also bind cooperatively with GR, and this also results in synergistic transcriptional regulation (156, 160). In vivo results indicate that Oct-1 binding to the MMTV promoter is strictly glucocorticoid dependent, consistent with a role for GR in remodeling of chromatin to promote access of other factors to the MMTV promoter, such as nuclear factor 1 (NF-1) and Oct-1 (161, 162).

On templates containing GREs and octamer sites, Oct-2A and GR can synergize transcriptionally (156, 163). However, when GR sites are not present, GR strongly inhibits transactivation by Oct-2A, while the reciprocal is not true. This effect does not seem to depend on interference with Oct-2A DNA binding, but more likely is due to competition for rate-limiting coactivator molecules (163). A minimal promoter driven by Oct-1, binding to an octamer element, can also be inhibited by GR in a hormone-dependent manner. This interaction seems to be mediated by a strong solution interaction between GR and Oct-1, resulting in inhibition of Oct-1 DNA-binding (164). This interaction, which seems to involve the POU-H domain of Oct-1, can be inhibited by GR binding to a GRE (156). While the mechanism is unclear, functional analyses indicate that androgen-mediated repression of the dehydroepiandrosterone sulfotransferase gene promoter requires Oct-1 binding to the promoter. Since there seems to be no direct AR binding to this promoter, AR-Oct-1 interactions may be important (165).

Oct-1 can interact with several other DNA-binding factors, including Sp1 (166, 167), AP-1 (168, 169, 170), CCAAT/enhancer binding protein-{beta} (C/EBP{beta}) (171), NF-1 (172, 173), and MEF2 (174). Oct-1/2 interactions with heterologous DNA-binding proteins can either result in transcriptional synergism, as exemplified by the AP-1/Oct-1 interaction (168, 169, 170), or interference as exemplified by C/EBP{beta} inhibition of Oct-1 function (171).

Interaction of the high mobility group protein HMG2 with the POU domains of Oct-1 and Oct-2 leads to cooperative binding (175). The related HMG I(Y), a known octamer-binding protein, functions as a coactivator with Oct-2A but not Oct-1 for HLA-DRA transcription. Apparently HMG I(Y) facilitates the binding of both Oct-1 and Oct-2 to octamer sites (176) by a mechanism that involves interactions between HMG I(Y) and the POU domains of Oct-2 and Oct-1. However, synergistic transactivation of the HLA-DT gene is only found with Oct-2 and seems to depend on sequences within the C terminus of Oct-2 (177).

b. Interactions with coregulator proteins.
The B cell-specific coactivator OCA-B (also referred to as OBF-1 and Bob-1), a 256-amino acid, proline-rich protein with no close homologs in the database, interacts specifically with the POU domains of Oct-1 and Oct-2 and is required for B cell-specific activation of immunoglobin genes (178, 179, 180, 181, 182, 183). By interacting with Oct-1 and Oct-2, OCA-B can be recruited to a subset of octamer sites where it serves a coactivator function for transcription (184) (Fig. 2BGo). In part, a strong transactivation domain in the C terminus of OCA-B mediates this effect, while the N terminus contains the domain interacting with POU domains (185). This interaction is specific for Oct-1 and Oct-2 (Fig. 2BGo) and not observed for other POU domain factors such as Oct-3, Oct-6, and Pit-1. The specificity of the interaction of OCA-B with both the POU-S and the POU-H domains is determined by residues in the first helix of the POU-S domain (L6 and E7) and the end of the third POU-H domain helix (K155 and I159) (124, 184). Other residues in the POU domain, including L53 and N59 in the POU-S domain, have been shown to be important for the Oct-OCA-B interactions. However, these residues are conserved in POU domain factors that do not interact with OCA-B and are thus unlikely to determine the specificity of the interaction.

Immunoglobin promoters are strongly activated by OCA-B, whereas the histone H2B promoter is poorly activated, suggesting that there is some promoter-specific restriction of the ability of OCA-B to transactivate. Further specificity is observed on the immunoglobin promoters because OCA-B only seems to activate on sites in the proximal promoter but not on distally located enhancer sites (124). In addition, the Oct-1/Oct-2-OCA-B complex can only occur on octamer sites that contain an A in the fifth position of the octamer site, such as found in the classical octamer site (Fig. 2BGo). Position +3, +4, and +6 also appear to be important (124, 184). Apparently, OCA-B makes direct contacts with the DNA backbone in the middle of the octamer site, via its N terminus, in addition to POU domain contacts (186, 187). It has been proposed that OCA-B clamps the two POU subdomains together and at the same time bridges the DNA within the octamer motif at several positions. According to this model, OCA-B may help organize the conformation and stabilize the DNA binding of the Oct POU domain on DNA (124, 184). There is also evidence that binding of OCA-B to the Oct-1 POU domain/octamer complex induces a partial folding of OCA-B (186).

Another extensively studied coregulator is VP16. In human cells infected with the herpes simplex virus (HSV), viral gene expression is initiated with the viral protein VP16, which regulates the HSV immediate-early (IE) genes. VP16 does not bind DNA directly, but is instead recruited to IE promoters by its association with Oct-1 (87, 188, 189, 190, 191, 192) (Fig. 2BGo). Another cellular protein, host cell factor (HCF), is required for nuclear import of VP16 (193) and stabilization of the VP16-Oct-1 interaction (194, 195, 196, 197). After infection, VP16 first forms a complex with HCF, which then promotes its interaction with Oct-1 and binding to the TAATGARAT motif in the HSV IE promoters (Fig. 2BGo). The interaction of the Oct-1 POU domain and VP16 is selective and not observed with the highly homologous Oct-2 POU domain (198) (Fig. 2BGo). Selective residues, particularly a glutamic acid residue (E22) located on the surface of the second helix of the POU-H domain, are critical for the recognition by VP16 (199, 200). In the Oct-2 POU-H domain, the analogous residue is an alanine, thus precluding interaction with VP16. In addition, the interaction between Oct-1 and VP16 requires the 3'-end GARAT of the TAATGARAT motif to be intact for ternary complex formation (89, 90) (Fig. 2BGo). VP16 alters gene regulation by Oct-1 in two distinct ways. First, it provides a transactivation domain that is highly active on mRNA promoters where the Oct-1 transactivation domain is inactive, and second, it stabilizes the Oct-1-binding to VP16-responsive elements (184).

c. Interactions with basal transcription factors.
SNAPc (snRNA activating protein complex; also referred to as PTF) is a basal transcription factor, composed of several well defined peptides, that binds to promoter elements called the proximal sequence element (PSEs) in both RNA Pol II and Pol III snRNA promoters (84, 201, 202). Oct-1 binds to the enhancer or distal sequence elements (DSEs) in snRNA promoters, which leads to cooperative binding with SNAPc via a protein-protein interaction involving the POU-S domain of Oct-1, especially residue E7, and the C-terminal region of SNAP190, a major component of the SNAPc (203). The result is an enhanced transcriptional activation by Oct-1-mediated recruitment of SNAPc to the PSE of snRNA genes (204). The carboxy terminus of SNAP190 acts as an inhibitor of SNAPc DNA binding, but the interaction with the Oct-1 POU domain relieves the inhibition and promotes cooperative binding; SNAPc binds with 8- to 10-fold higher affinity when Oct-1 is bound to the DSE (205, 206).

The POU domains of Oct-1 and Oct-2 have been shown to interact with the TATA binding protein TBP (207). Since this interaction was in solution and not on DNA, its significance is somewhat unclear. However, this finding is consistent with the observation that Oct-1 and Oct-2 can facilitate preinitiation complex formation (208). In addition, the binding of Oct-1 and Oct-2 to the lipoprotein lipase promoter octamer site was shown to be stimulated by transcription factor II B (TFIIB), apparently via protein-protein interactions (209).

A yeast two-hybrid screen with the Oct-1 POU domain identified MAT1 as a POU domain-binding protein capable of binding to Oct-1, Oct-2, and Oct-3 POU domains. MAT1 is required for the assembly of cyclin-dependent kinase (CDK)-activating kinase (CAK), which is functionally associated with the general transcription factor IIH (TFIIH). The interaction between POU domains and MAT1 can target CAK to octamer binding factors and promote their phosphorylation. This interaction may play a role in the recruitment of TFIIH to the preinitiation complex (210).

d. Interactions with replication factors.
Oct-1 is involved in adenovirus replication where it binds to the highly conserved core origin and interacts via its POU domain with one of the three viral proteins required for replication, the precursor terminal protein (pTP) (122, 211, 212). The interaction seems to involve several regions of the pTP protein (213). The POU-H domain binds to the pTP-DNA polymerase complex (213), but more weakly than the intact POU domain. No binding is observed with the POU-S domain alone (214). Presumably, this interaction facilitates the recruitment of the pTP-DNA polymerase complex to the origin of replication. The stimulation of adenovirus DNA replication seems to depend on a property of the POU domain that is common to many POU domains because other POU domain factors, including Pit-1 and Tst-1, can also stimulate DNA replication (215).

3. Class III POU interactions.
a. Interactions with DNA-binding proteins.
Based on far-Western and mammalian two-hybrid experiments, all class III POU domain factors can interact with each other via the POU homeodomain (216). Consistent with this observation, Brn-2 exhibits highly cooperative homodimerization on a consensus nonoctamer binding site (103).

Sox/HMG proteins are transcription factors that contain Sry boxes that mediate binding to the minor groove of DNA. Such factors may frequently interact with POU domain factors that bind to the major groove of DNA. Tst-1 (Oct-6) can interact with HMG I(Y) in solution, which facilitates Tst-1-binding to an AT-rich site in the JC viral regulatory region (217). Tst-1 can also interact with the related HMG2 (218). Both Tst-1 and Brn-1 can cooperate with different Sox proteins in transcriptional activation. Specifically, Brn-1 can synergize with Sox-11 when both proteins are bound to adjacent elements; the synergism depends on the transactivation domains in both proteins. Such POU-Sox interactions may be important for a combinatorial code required for cell-specific transcriptional activation (219, 220).

b. Interactions with coregulator proteins.
Tst-1 can interact with the large T antigen from the human papovavirus JC virus, and this interaction seems to lead to transcriptional activation on the early and late promoters from the JC virus (101, 221, 222). The amino-terminal 82 residues of the large T antigen and the POU domain of Tst-1 mediate this interaction, but for a functional effect on transcription, the N-terminal transactivation of Tst-1 is also required. This functional effect is specific for Tst-1, because while Brn-1 is also capable of interacting with large T antigen, this interaction is not transcriptionally productive, apparently because of differences in the transactivation domains of Brn-1 and Tst-1 (223).

4. Oct-3 interactions.
a. Interactions with DNA-binding proteins.
Oct-3 can interact with the HMG-box protein Sox2 (224), and these two factors have been implicated in synergistic regulation of the fibroblast growth factor 4 (FGF-4) gene (49) and the embryonic stem cell coactivator gene UTF-1 (112). The synergism between Oct-3 and Sox2 on the FGF-4 promoter seems to be due to cooperative binding, which can be disrupted by increasing the distance between the two binding sites (109). In contrast, Sox2 functionally antagonizes Oct-3 on the ostepontin gene-regulatory region where its binding site is adjacent to that of Oct-3 (111). Oct-3 and Sox2 interact directly in solution, with the interaction interface being the POU domain and the Sox2 HMG domain (109).

b. Interactions with co-regulator proteins.
In embryonic stem cells Oct-3 transactivation is independent of the distance of the octamer site from the promoter, but in differentiated cells, Oct-3 transactivation from a distance seems to require stem cell-specific coactivators that bridge the remotely bound Oct-3 to the basal transcriptional machinery (225). While these factors remain to be identified, it is known that transactivation by Oct-3 is strongly stimulated by E1A in 293 and HeLa cells and depends on both the intact POU domain and the transactivation domain of Oct-3. This effect is specific because it is not found with Oct-1 and minimally with Oct-2. Direct binding of Oct-3 to E1A was also demonstrated (225). The Oct-3 POU domain can associate with two distinct domains on the E1A protein, CR1 and CR2, and it can also associate with the HPV E7 protein (226). These data suggest that viral oncoproteins have characteristics that mimic a stem cell-specific activity (226).

5. Conclusions.
From studies on the protein-protein interaction characteristics of POU domain factors several general conclusions can be drawn. First, most factors that interact with POU domain proteins have been found to interact with the POU domain rather than other parts of the proteins. This finding may be due, in part, to technical difficulties in identifying factors that interact with the transactivation domains outside the POU domain. However, these results indicate that the POU domain is not only a DNA-binding domain but also an important interface for regulatory proteins. Second, POU domain factors seem fairly promiscuous as to their ability to interact and synergize with diverse DNA-binding proteins bound to the same promoter. Most POU domain factors have been shown to interact in such a way with nuclear hormone receptors and HMG proteins. Third, while the POU domain is highly conserved, most POU domain interactions with coregulators are remarkably selective; apparently proteins are able to discern small differences in POU domain sequence (Fig. 2BGo). Thus, the E1A protein interacts strongly with Oct-3 but not Oct-1 or Oct-2. VP16 interacts strongly with Oct-1 but not Oct-2, Oct-3, or Tst-1 (Fig. 2BGo). OCA-B interacts selectively with Oct-1 and Oct-2, but not with Oct-3 or Tst-1 (Fig. 2BGo). This selectivity for protein-protein interactions by POU domains is in stark contrast to the lack of selectivity in DNA-binding preferences. Fourth, the association of a coactivator with a POU domain factor may be stabilized by a simultaneous direct interaction of the coactivator with DNA, such as that found with Oct-1/2 and OCA-B (Fig. 2BGo). Fifth, some POU factor-coactivator complexes can only form on selective DNA sites, perhaps in part due to direct DNA contacts of coactivators or due to DNA-induced conformational changes of the POU domain as has been described with OCA-B/Oct-1 and VP16/Oct-1 interactions (Fig. 2BGo). For example, the Oct-1/2-OCA-B interaction requires an "A" in position 5 of an octamer site (ATGCAAAT) (184), and the Oct-1-VP16 interaction requires the GARAT part of a TAATGARAT site for ternary complex formation (90). Sixth, additional functional specificity is provided by sequences outside the POU domain, such that two distinct POU domain factors may be capable of physical interaction with a coregulator, but only one of the interactions will result in transcriptional activation, apparently because of specific sequences in the transactivation domain. Finally, while some protein-protein interactions involve only one of the subdomains of the POU domain, e.g., the interactions of VP16 with the POU-H domain and SNAPc with the POU-S domain, others such as OCA-B interact simultaneously with the POU-S and the POU-H domains.

D. Mechanisms of transactivation
1. Transactivation domains.
While the N- and C-terminal domains in POU domain factors are thought to act as transactivation domains, little is known about their structures. These domains, which frequently contain polymeric amino acid stretches, are poorly conserved between different POU domain factors, consistent with the idea that they confer specificity to the action of different factors, often binding to the same DNA element. The N terminus is serine/threonine rich in Pit-1, alanine/glycine rich in Tst-1, proline rich in Oct-3, and glutamine rich in Brn-2. The N- and C-terminal transactivation domains of Oct-2 have been most extensively characterized (227, 228, 229, 230). The N terminus contains an 18-amino acid glutamine-rich motif, and the C terminus contains a proline-rich domain (231). The function of these motifs is enhanced when they are combined or reiterated independently (229, 232), suggesting that they have a modular structure. The Oct-1 transactivation domain appears to be fairly specific for snRNA promoters and has less effect on mRNA transcription, in contrast to the Oct-2 transactivation domains that are active on mRNA promoters (184, 228).

Studies on Oct-3 suggest that transactivation domains may work in a cell-specific manner (233). The N-terminal sequence of Oct-3 fused to the GAL4 DNA-binding domain can transactivate in all cell types tested, whereas the C-terminal sequence shows cell-type specificity in the same assay. This cell type specificity of the C-terminal domain seems to be imposed by the Oct-3 POU domain because it functioned in all cell types tested when the C terminus was fused to Pit-1 POU domain. It is further possible that the transactivation specificity is regulated by phosphorylation as the variation in transactivating ability of the C-terminal domain correlates with differences in the phosphorylation status of Oct-3.

Studies on Pit-1 have defined additional specificity conferred by the N-terminal transactivation domain (234). Thus, synergistic activation by Pit-1 and the estrogen receptor on a PRL reporter construct requires an N-terminal 25-amino acid domain of Pit-1 that is not required for analogous synergism on the GH promoter. This synergy, which depends on two of three tyrosine residues, spaced by six amino acids, is preferentially used on sites where Pit-1 binds as a monomer instead of a dimer. These data suggest that the DNA-binding site, dictating whether Pit-1 binds in dimeric or monomeric form, can alter the segment of the N terminus used as a synergy domain.

The N terminus of Brn-2 contains a glutamine-rich region that is mainly encoded by CAG residues (36). Similar nucleotide triplets that can be expanded are found in disease genes causing fragile-X syndrome and myotonic dystrophy (235). It is unknown whether the same mutational mechanism might involve the Brn-2 gene. A polyglutamine tract-binding protein, PQBP-1, capable of binding to the polyglutamine tract of Brn-2 and other such tracts of triplet repeat disease, has been identified (236). While the function of this novel protein is unknown, it is known to be localized in the nucleus and capable of interfering with transactivation by Brn-2.

The N-terminal sequences of POU factors can also mediate repression, including regions from the N terminus of Oct-2 (237, 238, 239, 240, 241, 242). Oct-3 binds to the promoter of Rex-1, which encodes an acidic zinc finger protein expressed at high levels in embryonic stem cells (108). Depending on the cellular environment, Oct-3 can either activate or repress the promoter, with distinct domains within the N terminus being responsible for repression and activation.

2. Transactivation and chromatin.
Gene expression is influenced by chromatin structure (243). Histones are regulated by acetylation of lysines in their N-terminal tails, allowing changes in histone conformation generally associated with increased access of DNA-binding proteins to promoters. In contrast, histone deacetylation is coupled to transcriptionally silent chromosomal domains. Many transcriptional coactivators possess intrinsic histone acetylation activity that is targeted to specific genes through interactions with DNA-binding proteins. Repressors, on the other hand, can recruit deacetylating enzymes to promoters, either directly or via the corepressor mSin3 (244, 245).

N-CoR, originally identified as a nuclear receptor corepressor (246), has been reported to bind to the homeodomain of Pit-1 and actively suppresses transactivation by Pit-1. This suppression is dependent on mSin3, SAP30, and histone deacetylase (153, 247). N-CoR may be limiting under certain conditions, and its binding to other transcription factors, such as nuclear hormone receptors, may indirectly activate transcription by Pit-1. The Pit-1 POU domain also associates with the coactivator complex consisting of CBP/p300 and P/CAF, both of which possess histone acetylase activity (152, 153) (Fig. 3Go). Thus, the transcriptional activity of Pit-1 may be mediated by the competing binding of complexes responsible for either acetylation or deacetylation, resulting in activation or repression (Fig. 3Go).

3. Conclusion.
Ultimately, it is likely that POU domain factors mediate their transcriptional activation by a combinatorial mechanism involving the N- and C-terminally located transactivation domains and coactivators/corepressors that interact with the POU domain (Fig. 3Go). Since the involved coregulators participate in gene activation by many different transactivators, it is likely that the unique transactivation domains of POU domain factors provide specificity to the activation (228).


    III. Development of the Hypothalamo- Pituitary Region
 Top
 Abstract
 I. Introduction
 II. Structure and...
 III. Development of the...
 IV. Expression and Function...
 V. Relevance of POU...
 References
 
The mature pituitary gland is composed of three distinct anatomical and functional entities (Fig. 4Go) (139, 248, 249, 250): 1) the anterior lobe with five major hormone-producing cell types that respond to neuropeptides released from hypothalamic nuclei carried to the anterior pituitary gland via the portal circulatory system; 2) the posterior lobe containing specialized astroglia referred to as pituicytes, and magnocellular neuronal axons projecting from hypothalamic nuclei, releasing vasopressin and oxytocin (OT); and 3) the intermediate lobe containing melanotropes and endorphin- secreting cells (Fig. 4Go).



View larger version (29K):
[in this window]
[in a new window]
 
Figure 4. A schematic representation of the adult pituitary gland. The three lobes of the pituitary gland, anterior, intermediate, and posterior, are indicated with the main hormone-producing cell types. The X label over somatotropes, lactotropes, and thyrotropes indicates that these cells do not develop in Pit-1-mutated individuals.

 
The hypothalamus and the anterior pituitary gland develop from distinct regions of the most anterior part of ectoderm primordia (251). Yet the development of these two organs is highly coordinated with neuronal projection arriving in the median eminence and the posterior pituitary gland at about the same time as these neurons start producing neuropeptides and receptors for these peptides become expressed in the anterior pituitary gland. Classical embryological experiments have suggested that the development of the anterior pituitary gland is dependent on inductive signals from the diencephalon (252, 253, 254, 255).

Most of the hypothalamus appears to arise from the neuroepithelium ventral to the hypothalamic sulcus while the preoptic region arises rostral to the optic sulcus (256). During development of the hypothalamus, between embryonic days 11 and 18 (e11 and e18) in the rat, periventricular neurosecretory cells become organized into several nuclei. These include the paraventricular (PVH), supraoptic (SO), arcuate, and periventricular nuclei (PVN). Development of the hypothalamus appears to progress in an outside-in gradient with the late-arising neurosecretory cell types occupying the most medial or periventricular zone of the hypothalamus. For descriptive purposes, four rostrocaudal levels of the hypothalamus can be described: preoptic, anterior, tuberal, and mammillary.

Within the hypothalamus, two distinct neurosecretory systems are organized (Fig. 5Go) (256). One system is composed of magnocellular neurons that project axons directly into the posterior lobe of the pituitary gland (Fig. 5Go). The magnocellular neuronal system includes neurons of the PVH and SON that synthesize OT and vasopressin (VP) and release these peptides into the posterior lobe of the pituitary gland. The second system is composed of parvocellular neurons that synthesize neuropeptides released into the pituitary portal circulation for regulation of the anterior pituitary gland (Fig. 5Go). In addition to the magnocellular neurons, the PVH also harbors parvocellular neurons that synthesize, among other neuropeptides, CRH and TRH for regulating the ACTH/adrenal axis and the TSH/thyroid axis, respectively. In the anterior periventricular and arcuate nuclei are parvocellular neurons that provide dopaminergic control of lactotrophs, and somatostatin/GRH control of somatotrophs. GnRH neurons originate in the olfactory placode and migrate during embryonic development to the hypothalamus (257). GnRH is synthesized in scattered neurons throughout the hypothalamus, especially in preoptic hypothalamic neurons. It is involved in the initiation of puberty and regulation of FSH and LH synthesis in the anterior pituitary gland and thereby regulates reproduction.



View larger version (48K):
[in this window]
[in a new window]
 
Figure 5. A schematic representation of the adult hypothalamo-pituitary axes. Two types of neurons convey signals from the hypothalamus to the pituitary gland. Magnocellular neurons located in the SON and PVN project their axons directly into the posterior lobe of the pituitary gland where they release vasopressin (AVP) and OT. In contrast, parvocelluar neurons, located in several nuclei, including the paraventricular and arcuate nuclei, project their axons to the median eminence. There, these neurons release pituitary regulating signal substances that are carried via the pituitary portal circulation into the anterior lobe of the pituitary gland. The crossed out neurons are those missing in Brn-2-mutated mice.

 
The anterior pituitary gland originates from an invagination in the oral ectoderm, referred to as Rathke’s pouch (Fig. 6Go). When this oral ectoderm makes contact with the overlying neuroectoderm, around e9 in the mouse, a cascade of signaling events is initiated that ultimately lead to the generation of the anterior and intermediate lobes of the pituitary gland containing six distinct hormone-producing cells by e17.5 (139, 258). These cell types are corticotropes secreting ACTH, melanotropes secreting MSH, gonadotropes secreting LH and FSH, thyrotropes secreting TSH, somatotropes secreting GH, and lactotropes secreting PRL.



View larger version (25K):
[in this window]
[in a new window]
 
Figure 6. A model describing pituitary gland development in the mouse. At e6.5 a part of the anterior ectoderm is already specified to form the hypothalamus and pituitary gland (not shown). At e10 the rudimentary Rathke’s pouch is marked by expression of the LIM homedomain factor, Plim (Lhx3), within a larger region of Pitx-1-expressing cells in the oral ectoderm. Presumably, these transcription factors are responding to signals, including Shh, which is excluded from Rathke’s pouch, and Wnt5a, BMP-4, and FGF-8 released from the diencephalon. Later, around e11, these signals set up a spatial pattern of transcription factor expression as indicated. As development proceeds, the appearance of hormone-expressing cells appears in a temporal- and spatial-specific manner that corresponds to the pattern of transcription factor expression. The first hormone-producing cells to appear, around e12, are the Pit-1-independent rostral tip thyrotropes (Tr) and corticotropes (C). Later, around e15, gonadotropes (G) and caudomedial thyrotropes (T) appear in a ventral location, and somatotropes (S) and lactotropes (L) appear more dorsally in a caudomedial location. At that stage, corticotropes are widely distributed in the anterior pituitary gland. Somatotropes, lactotropes, and caudomedial thyrotropes form in a location that corresponds to the expression domain of Pit-1. For illustrative purposes the patterns of hormone-producing cells are shown as strictly demarcated segments, but in reality there are no strict boundaries in the developing gland. Later, in the adult anterior pituitary gland, hormone-producing cells are intermixed.

 
These cell types form in a spatial-specific manner earlier in mouse pituitary ontogenesis, around e10.5–14.5, during a period of active cell proliferation (Fig. 6Go). Thyrotropes and gonadotropes are most ventral and somatotropes and lactotropes reside in a dorsal position (139, 258) (Fig. 6Go). During this stage of pituitary organogenesis, at e15, corticotropes are distributed throughout the anterior pituitary gland (258, 259, 260). The spatial restricted organization of thyrotropes, gonadotropes, somatotropes, and lactotropes may be regulated by gradients of signaling molecules, including sonic hedgehog, bone morphogenic proteins (BMPs), FGFs, and Wnts, which in turn create overlapping spatial gradients of transcription factors (Fig. 6Go). Thus, the transcription factors Nkx3.1, Six3, and Prop1 are expressed predominantly dorsally; Isl1, Brn-4, P-Frk, SF-1, and GATA-2 are expressed predominantly ventrally; and Pit-1 is expressed in an intermediary location, the site where lactotropes and somatotropes initially appear (248, 249).

BMP4 signaling from the ventral diencephalon seems to be important for the early formation of the Rathke’s pouch and seems to be linked to the induction of the transcription factor Isl1 (261). Later, FGF8 released from the ventral diencephalon and BMP2 released from mesenchymal cells ventral to the pouch and within the pouch seem to be important for patterning and cell proliferation (261, 262, 263). Embryological studies in the rat suggest that the programs required for specification of all hormone-producing cell types is completed as Rathke’s pouch closes (around e12 in the mouse).

Concurrent with the appearance of the anterior and intermediate lobes of the pituitary gland, formed from the ventral and dorsal aspects of Rathke’s pouch, respectively, the posterior gland develops from neural ectoderm (Fig. 6Go).


    IV. Expression and Function of POU Domain Factors in the Neuroendocrine System
 Top
 Abstract
 I. Introduction
 II. Structure and...
 III. Development of the...
 IV. Expression and Function...
 V. Relevance of POU...
 References
 
A. Pit-1
1. Pit-1 proteins. The Pit-1 gene encodes a 33-kDa protein composed