Brn-5 Is a Divergent POU Domain Factor Highly Expressed in Layer IV of the Neocortex*

We have identified rat cDNA clones that encode a POU domain protein, referred to as brain-5 (Brn-5). During embryogenesis in the rat, Brn-5 is widely expressed with highest levels in the developing brain and spinal cord from embryonic day 12.5. In the adult, Bm-5 mRNA is most abundant in the brain, where it is diffusely expressed with the exception of an enrich- ment in layer IV of the neocortex. However, Brn-6 is also found in multiple adult tissues outside the central nervous system, including kidney, lung, heart, adrenal, skin, testis, and anterior pituitary. This expression pattern contrasts with that of most other POU domain genes that are expressed predominantly in the developing nervous system and are progressively restricted to discrete regions of the brain. The predicted amino acid sequence of Brn-5 is considerably divergent from previously described POU domains and thus defines a new POU domain subclass (class VI). Consistent with its divergent sequence, the DNA-binding characteristics of Brn-6 overlap with, but are clearly distinct from, that of Oct-2. Although Brn-5 can bind to non-octamer sites, a random site selection indicates that its preferred binding site contains a variant octamer core motif. Fi-nally, we show that the amino terminus of Brn-5 con- tains a transactivation domain.

important developmental functions were described in several other species, including mammals (for review see Holland and Hogan (1988) and Kessel and Gruss (1990)).
A specific subclass of homeodomain proteins was identified by the cloning of factors that activate specific transcription units; the octamer-binding proteins Oct-1  and Oct-2 (Clerc et al., 1988;KO et al., 1988;Muller et al., 1988;Scheidereit et al., 1988) and the pituitary transcription factor Pit-1 (Bodner et al., 1988;Ingraham et al., 1988). Comparison of these factors with the nematode developmental regulatory gene unc-86 (Finney et al., 1988) revealed a conserved 69-to 72-amino acid long domain on the NH2-terminal side of their homeodomain. This novel domain, referred to as the POU-specific domain, is required in conjunction with the POU homeodomain for high affinity interactions with most DNA-binding sites. The POU-specific domain is joined to the POU homeodomain by a 14-to %-amino acid variable linker sequence, and together they are referred to as the POU domain . The four initially described members of the POU domain family are thought to exert critical roles in gene regulation and metazoan development (Fletcher et al., 1987;Bodner et al., 1988;Ingraham et al., 1988;Muller et al., 1988;Scheidereit et al., 1988;Tanaka et al., 1988;Finney and Ruvkun, 1990;Li et al., 1990;Corcoran et al., 1993). In addition, Oct-1 and several other POU domain factors are capable of regulating viral DNA replication in vitro (Verrijzer et al., 1990a(Verrijzer et al., , 1992a, and Pit-1 appears to regulate pituitary cell proliferation (Li et al., 1990).
Since the molecular cloning of the original POU domain genes, several additional members of this family have been isolated (for review see Rosenfeld, 1991;Ruvkun and Finney, 1991;Scholer, 1991;. These additional POU domain proteins have permitted a classification of existing POU genes into five subclasses based on similarities in the sequence of the POU domain, including the linker region, and weaker homologies of the NH2-terminal domains (He et al., 1989;Hara et al., 1992). Although no unifying rule for expression pattern has emerged, examination of the developmental expression of POU domain genes in the mammalian nervous system has revealed that all known family members, except Oct-314 and S k n -l a l i are expressed during development ofthe nervous system (He et al., 1989; for review see Rosenfeld (1991) and Treacy and Rosenfeld (1992)). Thus, transcripts of the Oct-1, Oct-2, Pit-I, Brn-1, Bm-2, Brn-3.0, Brn-4, and l3t-1 genes are detected in the developing neural tube of the rat at embryonic days 9 to 12. During early stages, expression is diffise, and, in contrast to most classic homeodomain genes, the region of expression includes the developing midbrain and forebrain. During later embryological stages, and in the adult, expression of most of these genes becomes completely restricted to subregions of the brain (He et al., 1989Hatzopoulos et al., 1990;Monuki et al., 1990;Stoykova et al., 1992 l y s leu asp i l c thr pro l y s ser ala g l n l y s leu l y s pro val leu g l u l y s t r p ala leu thr ala thr g l u g I y pro ala tyr scr g l n scr ala i l c c y s arg phc glu l y s asn phc Iys i l e arg arg leu scr leu g l y leu thr g l n thr g l n Val g l y g l n g l u g l u i l e arg g l u phe ala his thr pm ser leu arp glu arp ply ile am leu l y s arg l y s arg arg thr scr phc thr pro g l n ala i l e gly gly glu pm ST 1. Nucleotide sequence and predicted amino acid sequence of Bm-6. The sequence shown includes in-frame stop codons 5' and 3' (designated ****) to the open reading frame. The POU-specific and the POU homeodomains are bored. The two direct repeats of 7 amino acids in the NH2 terminus are underlined. The numbers on the right refer to amino acids with the initial in-frame methionine assigned number 1.

POU Homeo Domain
and Young, 1992;Mathis et al., 1992;Gerrero et al., 1993). Because many POU domain genes were initially identified using a polymerase chain reaction approach with a common set of oligonucleotides, the divergence of this family may have been underestimated. Using a modified PCR-basedl approach, we have cloned cDNAs encoding a new POU domain protein, Brn-5, which has a prominent neuronal expression and is characterized by a highly divergent POU domain.
EXPERIMENTAL PROCEDURES PCR Cloning of POU Domains-Poly(A+) RNA from rat anterior pituitary glands was used as a template for cDNA synthesis with random hexamer primers and reverse transcriptase (Superscript, Life Technologies Inc.) according to instructions from the vendor. This cDNA was used as a template in PCR reactions with degenerate primers to con- The abbreviations used are: PCR, polymerase chain reaction; e, embryonic day; CMV, cytomegalovirus; SAAB assay, Selected And Amplified Binding site assay; bp, base paids); kb, kilobase(sl ACCA(GATC)AC-3']. The conditions of the PCR were as follows: 30 cycles at 94 "C for 1 min, 50 "C for 1 min, and 72 "C for 3 min. The PCR products were treated with restriction enzymes XhoI or BstXI and sizefractionated on a 2% agarose gel. Gel slices corresponding to full-length products were cut out, and the DNAwas isolated and used as a template for PCR with the same primers. The products from these reactions were cloned into a plasmid vector and sequenced.
Complementary DNA ClonineOne of the POU domains obtained in the screen described above was Bm-5. This fragment was 32P-labeled by random priming (Life Technologies Inc.) and used to screen a rat anterior pituitary cDNA library (Ingraham et d . , 1988). The cDNA clones obtained in this screen were sequenced on both strands as previously described (Andersen et al., 1993).
Messenger RNA Analyses-Northern blot analysis was done as previously described (Yu et al., 1991). The probe was a random primed 32P-labeled DNA fragment corresponding to the coding region for amino acids 39 to 189. RNase protection assays were done as previously described (Yu et al., 1991). The probe was a 32P-labeled cRNA corresponding to the coding region for amino acids 186 to 281.
In Situ Hybridization--Rat embryos and adult brain were treated as previously described Andersen et al., 1993). The probe that was used in most experiments was a 36S-labeled sense or antisense RNA corresponding to the coding region for amino acids 39 to 186. In some experiments, we used a 282-bp antisense probe containing 167 bp of 5'-untranslated sequence in addition to the first 115 bp of the coding region. Similar results were obtained with both antisense probes.  FIG. 2. Amino acid sequence comparison of mammalian POU domains. Known mammalian POU domains were assigned to classes I to V and compared as previously described (Rosenfeld, 1991). Bm-5 was placed in a new class (VI). The sequences are organized from left to right, and the lower halfis continuous from the top. The numbering systems for the POU-specific domain and the POU homeodomain are according to Assa-Munt et al. (1993) and Laughon (19911, respectively. Highly conserved residues are inside black and white boxes. Residues that are different in Bm-5 compared to most other POU domains are indicated with anasterisk. The four a helices in the POU-specific domain and the three predicted a helices in the POU homeodomain are indicated below. r indicates rat; h, human; and m, mouse. Other names that have been used for these mammalian POU domain factors are (for references see Wegner et al. (1993)): GHF-1 (for Pit-1); OTF-1, OBP100, NFIII, and NF-A1 (for Oct-1); OTF-2 and NF-A2 (for Oct-2); Oct-11 (for Skn-la); N-Oct-3 (for Bm-2); RHS2 (for Bm-4); Oct-6 and SCIP (for Tst-1); Bm-3 (for Bm-3.0); Oct-3 and Oct-4 (for Oct-3/4); Emb (for Bm-5).

Bm-5 Is Enriched in Neocortex
Protein-DNA-binding Assays-The proteins used for gel-shift analyses were in vitro translated or expressed in bacteria using a T, promoter, as previously described (Ingraham et al., 1990;Andersen et al., 1993). For the SAAB assay, we expressed Bm-5 holoprotein as a glutathione S-transferase fusion. The Bm-5 protein was purified by glutathione agarose affinity chromatography (Smith and Johnson, 1988) followed by cleavage with factor X. We observed no difference in binding specificity between proteins prepared by the three different methods. Gel mobility shift assays were done as previously described (Andersen et al., 1993). All binding sites were 32P-labeled to similar specific activity with T4 polynucleotide kinase. The sequences and origin of the binding sites have been previously described (Mathis et al., 1992) except for En, 5'-AAGGGGATCCAAATGTCAATTAAATATCAA-3'; POMC CE2, 5'-TCCTCAITAGTGATA'I"Il'ACCTCCAAATGC-3'; and I12 Oct, 5'-TTTGAAAATATGTGTAATATGTAAAACATlTFG-3'. The SAAB assay was performed as previously described (Blackwell and Weintraub, 1990). The sequence of the template was: 5"CGATGAA'ITCC-TAAGCGCATNNNNNNNNGAGCTCAGATC TC-3'. The sequences of the primers used to amplify the template were: 5"CGAT-GAAITCCTAAG-3'(sense) and 5'-ACGAGATCTGAGCTC-3' (antisense). The conditions of the PCR were as follows: 25 to 35 cycles of 94 "C for 45 s,48 "C for 2 min, and 72 "C for 30 s. After three rounds of selection, the template was sequenced using a 32P-labeled sense primer as previously described (Blackwell and Weintraub, 1990).
Cell Culture and Dansfections-CV-1 cells were grown and transfected with the calcium phosphate method as previously described (Andersen et al., 1993). CMV Lex A Bm-5 contains the coding region for amino acids 1 to 146 fused in-frame with the DNA-binding domain of LexA and placed downstream of the CMV enhancer/promoter as previously described (Ingraham et al., 1990). CMV LexA Pit-1, 2X LexA -36 luciferase and -36 luciferase plasmids have been described before (Ingraham et al., 1990).

RESULTS
Cloning and Characterization of Brn-5 cDNh-Degenerate oligonucleotides representing all possible codons in two different 9-amino acid conserved regions in Oct-1, Oct-2, Pit-1, and Unc-86 were initially utilized as primers for PCR to identify novel POU domain genes (He et al., 1989). However, with the subsequent cloning of the Oct-3/4 gene (Okamoto et al., 1990;Rosner et al., 1990;Scholer et al., 1990a), it became evident that the 5' region on which one of the degenerate oligonucleotide primers was based, located in the POU-specific domain, was not fully conserved. This observation raised the possibility that additional POU domain genes might have escaped detection by the initial PCR approach. Therefore, we used an oligonucleotide to another highly conserved region in the middle of the POU-specific domain (5'-oligonucleotide) in combination with an oligonucleotide to the highly conserved third helix of the POU homeodomain (3'-oligonucleotide). The cDNA was generated from rat anterior pituitary mRNA using random hexanucleotides to increase the chance of detecting POU domain transcripts with long 3"untranslated sequences. Because Pit-l is a highly abundant POU domain transcript in the pituitary, we excluded Pit-1 transcripts by treating the amplified material with either restriction endonuclease XhoI or BstXI, both of which cleave in the Pit-1 POU domain. The amplified material was then size-fractionated by agarose gel electrophoresis, and DNA corresponding to a full-length PCR fragment was isolated and reamplified. These products were subsequently cloned into a plasmid vector and analyzed by dideoxynucleotide sequencing.
We used this method to isolate several copies of a novel POU domain, referred to as Bm-5, which was selected for further analysis. The Bm-5 PCR fragment was used to screen a rat pituitary cDNA library, yielding three overlapping clones of 4.6 kb, 4.7 kb, and 2.1 kb denoted A Z A P Bm-5A, -B, and -C, respectively. All three clones contained an identical open reading frame of 900 bp (Fig. 1). The translation start site was assigned based on an in-frame stop codon 5' to the putative initiating ATG, as well as homology to a consensus Kozak sequence (Kozak, 1984). All three clones contained poly(A) tracks at their 3' end. Clones A and B were similar in that both contained 2.9 kb of 3'-untranslated sequences, whereas clone C contained only 750 bp of 3"untranslated sequences, suggesting the use of an alternative poly(A) signal site. Although heterogeneity was detected in the 5'-untranslated regions, this did not alter the open reading frame. Brn-5 is a single copy gene (data not shown), and we have localized the gene to mouse chromosome 15 and human chromosome 12 (Xia et al., 1993).
The POU domain of Bm-5 is unusual because it has significant alterations in several predicted amino acids that are invariant or exhibit only conservative changes in previously described mammalian POU domains (Fig. 2). In the POU-specific domain, 11 of the highly conserved amino acids are different, and 6 of these changes are nonconservative. In the POU homeodomain, seven near-invariant amino acids are altered in Bm-5. Five of the changes in the POU homeodomain are nonconservative. The 20-amino acid linker region of Bm-5 bears no homology to any of the previously described POU proteins.
Because of this high degree of divergence, we assigned Bm-5 to a new subclass (class VI) of the POU domain gene family (Fig.  2). Indeed, two highly similar sequences from zebrafish and mouse, referred to as pou[cl (Johansen et al., 1993) and Emb (Okamoto et al., 19931, respectively, were recently reported. The predicted amino acid sequences of pou[c] and Emb are 70% and 97% homologous to that of rat Bm-5, respectively. The pou[cl coding region contains a 302-amino acid NHz-terminal extension that is not in Emb or Bm-5. Other structural features of Bm-5 include an unusually short COOH terminus of only 10 amino acids. The N H z terminus, which is 144 amino acids long, bears no obvious homology to previously described POU domain genes. However, it features an over-representation of prolines (17%), a characteristic which has been associated with trans-activation domains in several other transcription factors (Mitchell and Tjian, 19891, including Oct-314 (Imagawa et al., 1991). In addition, the NHz terminus contains a 7-amino acid sequence, NAQGQVI, that is repeated twice (Fig. 1).
Expression of Brn-5 mRNA during Development and in the Adult-RNA blot analysis using RNA from multiple tissues and cell lines was used to determine the size and distribution of Bm-5 transcripts (Fig. 3). A major transcript of 5.5 kb was observed in several somatomammotroph pituitary cell lines (GC, MMQ, 235-1) and in a thyrotroph tumor (TtT 97). Upon longer exposure, low level expression was found in embryonic cell lines F9 and P19 and in the corticotroph cell line AtT-20, but no expression was detected in HeLa, HL-60, or CV-1 cells (data not shown). In addition, minor bands of 2.8 kb and 2.1 kb were found in pituitary cell lines and embryonic cell lines, respectively. Although the nature of these minor transcripts is unclear, the 2.8-kb mRNA is consistent with the use of the alternative poly(A) signal site found in A Z A P Bm-5C.
To further determine the pattern of Brn-5 expression, we performed RNase protection assays on RNAs from several different rat adult and embryonic tissues in addition to cell lines (Fig. 4, A and B). In the adult, Brn-5 is most highly expressed in the brain but transcripts are also readily detected in kidney, lung, heart, skin, adrenal, and placenta. Low level expression was observed in spleen, muscle, liver, anterior pituitary, testis, and ovary. During development, expression was readily detected in the head a t embryonic day 15, with levels decreasing gradually until postnatal day 10. On embryonic day 17, expression was detected in brainstem, cortex, and hypothalamus, with highest levels in the cortex. The low level of expression in testis appears to be developmentally controlled, being highest during the prepubertal stage (Fig. 4B). Together, these data suggest that Bm-5 is expressed a t highest levels in the central nervous system, but at lower levels in many different organs.
We utilized in situ hybridization to localize the expression of Bm-5 more precisely both during development and in the adult brain. A series of rat embryos ranging from stage e9.5 through e15.5 were probed with sense and antisense Bm-5 RNAprobes.
No expression of Bm-5 was detected before e12.5 (Fig. 5, left panel, and data not shown). On e13.5 and e15.5, expression of Bm-5 was detected diffisely throughout the developing brain and spinal cord (Fig. 5, middle and right panels). Because Bm-5 signal could be detected in whole adult brain by RNase protection, it was likely that Bm-5 was either expressed diffusely or, alternatively, restricted with very high levels in expressing cells. To distinguish between these possibilities, we camed out in situ hybridization studies on regularly spaced coronal brain sections. Expression was diffise throughout the brain, and, in general, the signal correlated with cell density, including regions such as the hippocampus and cerebellum. treatment with RNase, the products were size-fractionated on a 5% denaturing polyacrylamide gel and analyzed by autoradiography. The numbers on the right show the size of the protected fragment in rat tissues (290 bp) as judged by the migration of a labeled 1-kb DNA ladder (Life Technologies Inc.). Hybridization to RNA of mouse origin gives rise to two fragments, 160 and 100 bp long, apparently due to sequence divergence between mouse and rat in this region of the Brn-5 gene. B50 and BI09 are neuroblastoma cell lines. CA 77 and GC are thyroid C cell and somatotroph cell lines, respectively. TtT 97 is a mouse thyrotroph tumor. d l 0 Head refers to a brain of a 10-day-old rat. B, 20 pg of total RNA from the indicated sources were hybridized to a Bm-5 (top panel) and P-actin probes (lower panel) and analyzed as described above. The one clear exception was a higher expression level found in neocortex, and, within the cortex, the signal was most intense in the inner granular layer (layer W, Fig. W). Observation under high magnification light field showed that silver grains were localized over neurons (data not shown). No signal was observed when we used a sense Brn-5 riboprobe (data not shown). An RNase protection assay using RNA from various brain regions was used to independently confirm the expression pattern showed by the in situ hybridization studies. This experiment revealed that Bm-5 was expressed in all brain regions, with highest expression in the neocortex (Fig. 6B).
Taken together, the in situ hybridization and RNase protection data indicate that Bm-5 mRNA has widespread distribution in adult brain, and that the expression is enriched in the neocortex.

DNA-binding Properties of Brn-S-To test whether the diver-
gent sequence of the Bm-5 POU domain led to different DNAbinding properties, we expressed Bm-5 both in bacteria and in uitro using rabbit reticulocyte lysate. Binding of the expressed protein was tested in the gel-mobility shiR assay using a series of AT-rich candidate binding sites for POU proteins or classic homeodomain proteins (Fig. 7). Bm-5 binding to the following elements was easily detected in order of decreasing affinity: HSV Oct, CRH Oct, POMC DE2, Ftz, pOct, and H+O+. Unexpectedly, Brn-5 bound poorly to an element, H-O+, where the heptamer binding site was mutated. This suggests that, in contrast to classic octamer-binding proteins, the Bm-5 binding site overlaps with both the octamer and heptamer sequences. We could not detect binding to unrelated sequences, such as a thyroid hormone response element and a site that binds helixloop-helix proteins (E box; data not shown). This binding pattern is different from that exhibited by Oct-2, which prefers all sites that contain a core octamer element: H'O', H-O+, pOct, and I12 Oct (data not shown). Thus, consistent with the divergent sequence of the Brn-5 POU domain, its binding preference appears to be distinct from that of Oct-2. Bm-5 (upper panel) and a rat p-actin probe (lower panel ). After treatment with RNase, the products were run on a 5% denaturing polyacrylamide gel, and the gel was autoradiographed. The protected fragments are indicated on the right side.
Alignment of the high affinity binding sites for Bm-5 revealed a potential consensus sequence: 5'-GCATNN(N)TAAT-3' in the three highest affinity binding sites, whereas the same sequence was not found in any of the lower affinity sites ( Fig.  8A and data not shown). These data suggest that the DNAbinding domain of Bm-5 may bind to bipartite sequences, separated by 2 or 3 base pairs. To test this hypothesis, we carried out a mutational analysis of the high affinity CRH site and tested binding of Bm-5 by the gel-mobility shift assay (Fig. 8B 1. This analysis revealed that mutation of a -C-5' to the consensus had no effect on binding (Fig. 8B, lane 21, whereas double mutations through the -GCAT-sequence on the leR (Fig. 8B,   lanes 3 and 4 ) and the -TA-on the right (Fig. 8B, lane 7) completely obliterated binding. Although two separate mutations of the 3-bp sequence between the two half-sites decreased binding affinity, the effect of these mutations was less dramatic (Fig. 8B, lanes 5 and 6). The relaxed sequence requirements in this region of the binding site are consistent with two critical regions of Bm-5 binding sites separated by 3 base pairs.
The classic octamer site, 5'-ATGCAAAT-3' has been previously described as a bipartite site with the homeodomain contacting the -AAATsequence on the 3' portion of the site, and the POU-specific domain contacting the -ATGC-on the 5' side (Kristie and Sharp, 1990;Verrijzer et al., 1990bVerrijzer et al., , 1992bLaughon, 1991). The site described here for Bm-5 could be considered different because of a presumed separation of the two binding regions and the relaxed sequence specificity between the two halves. To test whether Bm-5 binds as a dimer to this site, we took advantage of the approach described by Hope and Struhl(1987). The binding of Bm-5 holoprotein and Bm-5 glutathione S-transferase fusion protein to the CRH site was tested using the gel-mobility shift assay. Bm-5 holoprotein (Fig. 8C, lane 2) and Bm-5 glutathione S-transferase fusion protein (Fig. 8C, lane 4 ) form complexes with distinct mobility patterns. When the two proteins were mixed together, no complexes with intermediate migration were observed (Fig. 8C,  lane 3 ) . Similar results were obtained using the HSV octamer site and when we used co-translated products from templates encoding Bm-5 holoprotein and POU domain (data not shown).
These data suggest that Bm-5 binds as a monomer to both the HSV Oct and CRH sites.
The unusual binding site determined in these experiments suggested the possibility that Bm-5 might bind preferentially to non-octamer sites. To test this possibility more rigorously, we used the "selected and amplified binding site" (SAAB) assay (Blackwell and Weintraub, 1990) to identify a preferred site for Bm-5. The site in this study contained the 5'-GCAT-3' sequence followed by 8 random nucleotides. By the means of this assay we identified a preferred site: 5'-GCATATGATAAT-3' (Fig. 9A). Surprisingly, this site is very similar to the preferred site for Oct-1: 5'-(AlT)TATGC(APT)AAT-3', that was recently identified by a random binding site assay (Verrijzer et al., 1992b). Thus, although Bm-5 contains the most divergent POU domain yet described, it binds preferentially to a variant of an octamer binding site that is very similar to the optimal binding site for Oct-1 (Fig. 9B).
!lYanscriptional Activity of Brn-&We initially evaluated the potential transcriptional function for Bm-5 using a transient co-transfection assay in CV-1 and HeLa cells. Bm-5 was expressed under the control of the CMV enhancer/promoter along with a reporter plasmid where octamer sites or three copies of the CRH site were linked to a minimal promoter and a luciferase reporter gene. Bm-5 activated these reporters only 2-fold, perhaps due to interference by endogenous octamer-binding proteins, such as Oct-1, which can activate these reporters

FIG. 8. Identification of a consensus high affinity DNA-binding
site for Bm-6. A, alignment of highest affinity binding sites for Bm-5 identified in Fig. 7. In some instances, two possible alignments are shown. Sense (SI and antisense (AS) strands are indicated. B, mutational analyses of the CRH site. The wild type CRH oligonucleotide or the indicated mutant sites were radioactively labeled, incubated with Bm-5 protein, and analyzed by the gel-mobility shift assay. Bound ( B ) and free ( F ) probes are indicated on the right. C , Bm-5 binds as a monomer. A radioactively labeled CRH site was incubated in the absence of protein (lune 1 ), with Bm-5 holoprotein (lune 2 ) , with Bm-5 glutathione S-transferase fusion protein (lam 41, or with Bm-5 holoprotein and Bm-5 glutathione S-transferase fusion protein together (lane 3). Bound ( B l , Bm-5 glutathione S-transferase-DNA complex, and B2, Bm-5 holoprotein-DNA complex) and free ( F ) probes are indicated on the right.
(data not shown). To independently test whether Brn-5 contains a domain capable of transcriptional activation, a CMV LexA Brn-5 fusion plasmid containing the NH2 terminus of Brn-5 linked to a LexA DNA-binding domain, was co-transfected with a reporter plasmid where 2 LexA binding sites were linked to a minimal promoter and a luciferase reporter. Under these conditions, we observed a consistent &fold activation imprinted by the NH2-terminal Brn-5 information (Fig. 10,   black bars). This activation was similar to the activation by CMV LexA Pit-1 (Fig. 10) that contains the transcriptional activation domain of Pit-1, a POU protein known to be capable of activating both the growth hormone and prolactin genes, as well as to that of CMV LexA Brn-4 (Mathis et al., 1992). The activation was specific because a Lex fusion containing the NH2 terminus of Skn-li, which is incapable of DNA binding, did not activate in these same experiments (data not shown). The ac- tivation was site-dependent because a reporter plasmid without Lex sites was not stimulated (Fig. 10, white bars). These data suggest that Brn-5 can function as a transcription factor.

DISCUSSION
Defining additional POU domain proteins is of interest because at least three of these factors, Pit-l (Li et al., 1990),Oct-2 (Corcoran et al., 19931, and unc-86 (Finney and Ruvkun, 1990) have been genetically shown to determine cell fate or differentiated function, suggesting that there may be important roles for each of the diverse members of this family. We have identified a cDNA encoding a novel POU domain factor referred to as Brn-5, whose predicted amino acid sequence is highly diverged from other POU domain proteins, thus defining a new class of POU domain factors (He et al., 1989;Johansen et al., 1993;Okamoto et al., 1993).
The POU domain is a bipartite DNA-binding domain composed of a POU-specific domain joined by a short linker sequence to the POU homeodomain. The POU-specific domain contains four a helices, homologous to the helix-turn-helix motif in the h repressor (Assa-Munt et al., 1993;Dekker et al., 1993). The structure of the POU homeodomain is thought to be similar to that of the classic homeodomains with a cluster of basic amino acids at the N H 2 terminus and three a helices (Laughon, 1991;Rosenfeld, 1991;Scholer, 1991). In fact, POU domains are highly conserved, even across species including Drosophila (Johnson and Hirsh, 1990;Billin et al., 1991;Dick et al., 1991;Lloyd and Sakonju, 1991;Treacy et al., 1991Treacy et al., , 1992Prakash et al., 19921, Caenorhabditis elegans (Finney et al., 1988), Xenopus (Agarwal and Sato, 1991;Frank and Harland, 1992;HinMey et al., 1992;Whitfield et al., 1993), zebrafish (Matsuzaki et al., 1992;Johansen et al., 19931, chicken (Petryniak et al., 19901, and mammals (Rosenfeld, 1991;Ruvkun and Finney, 1991;Scholer, 1991;Hara et al., 1992;. Although the predicted structure of the Brn-5 POU domain probably conforms to these general rules, it has alterations of many amino acids that are conserved across all previously described mammalian POU domains. For instance, the serine residue that replaces a lysine at position 22 in the basic region of helix 1 of the POU-specific domain of Brn-5 creates a potential phosphorylation site for protein kinase A (Kennelly and Krebs, 1991). Similarly, Oct-3/4 contains a threonine at the same position. This is reminiscent of the basic region in the NH2 terminus of the POU homeodomain in which a phosphorylation site regulates the binding activity of l'it-1 (Kapiloff et al., 1991) and Oct-1 (Segil et al., 19911, raising the possibility that activity of the POU-specific domain of Bm-5 and Oct-3/4 might also be regulated by phosphorylation. Interestingly, Brn-5 has an alteration in the DNA recognition helix (alanine instead of threonine at position 46) of the POU-specific domain. However, this residue may not be directly involved in DNA recognition (Dekker et al., 1993). Two of the altered amino acids in the POU homeodomain, at position 25 (leucine instead of a basic residue) and 55 (threonine instead of a basic residue), are involved in phosphate backbone contacts in binding of classic homeodomains to their cognate site (Laughon, 1991). The NH2 terminus of Brn-5 contains a 7-amino acid sequence, NAQGQVI, that is repeated twice. Although its function is unknown, it is likely to be important because it is completely conserved in the mouse. Furthermore, in zebrafish, one of the repeats is unchanged while the other has two conservative changes. Because the NH2 terminus is a transactivation domain, it is tempting to speculate that this domain might be involved in protein-protein interactions that are required for transcriptional activation.
Brn-5 exhibits a pattern of expression that is qualitatively distinct from most other POU domain factors. The most striking feature of Brn-5 gene expression is the widespread distribution of Brn-5 transcripts, somewhat similar to that of Oct-1 Kambe et al., 1993). Whereas, Oct-314 and Tst-1 are expressed early in embryogenesis and in embryonic cell lines (Meijer et al., 1990;Okamoto et al., 1990;Rosner et al., 1990;Scholer et al., 1990aScholer et al., , 1990bSuzuki et al., 19901, Brn-5 is expressed at low levels in these cell lines and is not detected by in situ hybridization in the rat embryo until embryonic day 12.5. Therefore, high levels of Brn-5 expression appear to be associated with the appearance of more differentiated cell phenotypes. Similar to most POU domain genes, Bm-5 is initially diffusely expressed in the developing central nervous system. However, in contrast to neuronally expressed POU domain genes of class I11 and IV, Brn-5 mRNA expression does not subsequently become restricted to limited regions of the nerv-ous system. Instead, Brn-5 is expressed diffusely throughout the adult brain, and signal strength correlates with cell density in structures such as the hippocampal formation (Okamoto et al., 1993) and cerebellum. A clear exception to this pattern is focally enriched expression in the neocortex, especially layer IV. In this regard, Brn-5 expression is distinct from Brn-2 and Tst-1, which are expressed in isocortex layers 2-5 and 5-6, respectively. In layer IV, sensory inputs from the thalamus or adjacent cortical regions terminate on neurons that connect to adjacent cortical laminae (Schmitt et al., 1981). Interestingly, we also observed enriched expression in the piriform cortex, which receives olfactory input. Therefore, Bm-5 joins the growing list of POU domain genes that are expressed in neurons involved in sensory pathways. In C. elegans, unc-86 specifies several sensory neuronal pathways (Finney and Ruvkun, 1990) and in mammals the Bm-3 gene is highly expressed in sensory ganglia (He et al., 1989;Gerrero et al., 1993). Furthermore, various products of the Oct-2 gene (Stoykova et al., 1992) and Brn-2 (He et al., 1989) are expressed in the olfactory pathway. Therefore, it appears that several POU domain genes may have important roles in the development of neurons involved in different aspects of processing sensory information in the mammalian nervous system. Brn-5 is also expressed in several organs outside the nervous system. Expression of Brn-5 in testis becomes lower after puberty, suggesting that Bm-5 is expressed in non-germ cells or in immature spermatogonia.
Although POU homeodomain DNA recognition is thought to be similar to that of the classic homeodomain, high affinity DNA binding to cognate sites also requires the POU specific domain Ingraham et al., 1990;Kristie and Sharp, 1990;Verrijzer et al., 1990b;Imagawa et al., 1991), which makes direct DNA contacts (Aurora and Herr, 1992;Verrijzer et al., 1992b). Like the POU domain, the DNA recognition sequences for POU homeodomain proteins also appear to be bipartite. Oct-1 binds with highest affinity to sites containing an octamer motif: 5'-ATGCAAAT-3' (for review see Laughon (1991)). The 3' half of this site ( -U T -, or a variant thereon is recognized by the homeodomain whereas the 5' half (-ATGC-) is recognized by the POU-specific domain (Kristie and Sharp, 1990;Verrijzer et al., 1990bVerrijzer et al., , 1992b. Oct-1 can also bind to a different type of sequence, the so-called TAATGARAT motif (Sturm et al., 1987); however, it binds to this site with lower affinity and may not require the POU-specific domain for this interaction (Verrijzer et al., 1990b). Because of the divergent sequence of the Brn-5 POU domain, it was of interest to determine whether Bm-5 preferred octamer DNA-binding sites. These experiments are especially relevant because the zebrafish homolog of Bm-5 was reported to represent the initial example of a non-octamer-binding POU domain protein (Johansen et al., 1993). Although our results indicate that Bm-5 is capable of binding to non-octamer sites containing the consensus sequence: B'-GCATNN(N)TAAT-3', a random binding site selection demonstrates that Bm-5 binds preferentially to a site containing a variant octamer element, surprisingly similar to that preferred by both Oct-1 (Verrijzer et al., 1992b) and a new class V POU domain protein.2 Therefore, our results are consistent with previous studies demonstrating that, although different POU homeodomain proteins may recognize distinct sequences, they generally have the ability to bind to certain sites containing the octamer motif (Singh et al., 1986;Sturm et al., 1987;Baumruker et al., 1988;LeBowitz et al., 1989;Poellinger and Roeder, 1989;Scholer et al., 1989;Meijer et al., 1990;Okamoto et al., 1990;suzuki et al., 1990;Hinkley et al., 1992;Mathis et al., 1992).