Two Novel Brain-specific Splice Variants of the Murine Cβ Gene of cAMP-dependent Protein Kinase*

We have previously characterized two murine cAMP-dependent protein kinase catalytic subunit genes, Cα and Cβ1. Targeted disruption of the Cβ1 promoter revealed two splice variants of the Cβ catalytic subunit gene (designated Cβ2 and Cβ3) that continue to be expressed. These variants arise from unique promoters and are brain-specific. Cβ2 is expressed in several discrete areas in the limbic system. These include the lateral septum, the bed nucleus of the stria terminalis, the ventral medial hypothalamus, and the amygdala. In the neocortex, expression is highest in cortical areas such as the prefrontal and insular cortex that are associated limbic structures. By contrast, Cβ1 is most highly expressed in the cortex and hippocampus and is also present in all non-neuronal tissues examined. Cβ3 is expressed at very low levels with wide distribution throughout the brain. Both the Cβ2 and Cβ3 variants are enzymatically active and induce gene expression in transient transfections with a cAMP response element-reporter construct. This activity is inhibited by protein kinase A regulatory subunits, the protein kinase inhibitor, and the chemical inhibitor H-89. We also demonstrate that Cβ1 is myristoylated at the amino terminus like the Cα isoform, but neither Cβ2 nor Cβ3 is myristoylated. The discrete expression of Cβ variants in the brain suggests specific functional roles in neuronal signaling.

Cyclic AMP-dependent protein kinase (protein kinase A (PKA) 1 ) has been shown to be the principal mediator of cellular responses to cAMP in animal cells. The inactive PKA holoenzyme exists as a heterotetramer of two regulatory (R) and two catalytic (C) subunits. The C subunit family consists of two characterized isoforms (C␣ and C␤) that have been described in murine (1), bovine (2), human (3), and porcine (4) tissues. The C␣ isoform is expressed ubiquitously, whereas C␤, although found in all tissues examined, is most highly expressed in the brain (5). A third isoform (C␥) is found only in primates and is expressed from a processed gene in the testis (3). The murine genome also contains a processed pseudogene related to C␣, but this pseudogene is not transcribed (6).
Although the C␣ and C␤ isoforms are 91% identical in amino acid sequence, recent work has revealed distinct biochemical properties that suggest unique functions (7). C␣ exhibited a 3-5-fold lower K m for certain peptide substrates and a 3-fold lower IC 50 for inhibition by PKI or RII␣ subunits than did C␤. In addition, holoenzyme containing RII␣ and C␤ is 5-fold more sensitive to activation by cAMP and has a higher basal activity in COS-1 cells than does holoenzyme containing C␣.
The C␣ subunit of PKA was one of the first examples of an N-myristoylprotein to be described (8). In the case of members of the Src family of protein-tyrosine kinases, this modification is necessary for targeting the protein to membranes and for normal function (9,10). N-Myristoylation of C␣ does not appear to serve a unique membrane targeting function as the protein is found in many cellular compartments. We have also demonstrated previously that a mutant form of C␣ that is not myristoylated has the ability to induce gene expression and modulate steroidogenesis (11).
A targeted disruption of the C␤ gene was created in which the promoter and translational start site were deleted, and mice carrying this mutation lacked the C␤ protein that we had previously characterized (12). We have designated this previously described isoform C␤1. Mice that lack C␤1 exhibit a decrease in hippocampal long-term potentiation in the Schaffer collateral-CA1 pathway. In addition, the C␤1 mutant mice lack both long-term depression and depotentiation in the Schaffer collateral-CA1 pathway. We have also shown that these mice lack long-term potentiation in the mossy fiber-CA3 pathway (13). However, Northern analysis of tissues from C␤1 knockout mice showed that the C␤ transcript was eliminated from all tissues of the C␤1 mutants except the brain, where a unique transcript remained. Western blots of brain regions probed with a C␤ antibody revealed a slightly more rapidly migrating band that remained in the knockout mice.
In this paper, we describe two novel splice variants of the murine C␤ gene. These two C␤ variants (designated C␤2 and C␤3) diverge from the original C␤1 sequence at the amino terminus and arise from the use of alternate promoters. Both the C␤2 and C␤3 variants are brain-specific and encode functional catalytic subunits that interact with R subunits and PKI. In contrast to the previously studied C␣ and C␤1 proteins, C␤2 and C␤3 are not myristoylated at the amino terminus. The C␤2 transcript is most highly expressed in limbic areas of the brain, whereas C␤1 is highly expressed in the hippocampus and more diffusely throughout the brain. Expression of C␤3 is very low and appears to be widely distributed in brain regions.

EXPERIMENTAL PROCEDURES
Oligonucleotide Probes-Oligonucleotides were made to correspond to unique sequences from the first exons of the three C␤ variants. Sequences are as follows: C␤1, 5Ј-AAGAAAGGCAGCGAAGTGGAGA-GC-3Ј; C␤2, 5Ј-GAGACATTGCCTGTCATCATGAAT-3Ј; and C␤3, 5Ј-C-CTGCTGGATCCAACATGGGCTTG-3Ј. In addition, oligonucleotides were generated corresponding to common sequence found in exons 2 and 3 of the C␤ gene: exon 2, 5Ј-GAGGGTTCTCCCATTTCC-3Ј; and exon 3, 5Ј-AGTAATGCTGGGCTTGAG-3Ј. These oligonucleotides were end-labeled with 32 P and used to probe Southern blots of genomic clones isolated in our initial characterization of C␤ (14).
Sequencing-Positive cDNA clones identified by screening a mouse brain cDNA library were sequenced using Taq Dye Primer (Applied Biosystems). These were aligned with known C␤1 sequence and used to generate oligonucleotide probes for mapping. A 3.2-kilobase EcoRI fragment from clone 15, which contains exons 1 of C␤2 and C␤3, was subcloned into pUC19. Automated sequencing was performed beginning with pUC19 forward and reverse sequencing primers followed by Taq DyeDeoxy Terminator (Applied Biosystems) sequencing using internal primers designed from previous sequence.
Construction of Riboprobe Vectors-Vectors to produce RNA probes for Northern blotting and in situ hybridization analysis were generated by PCR from cDNA clones, followed by subcloning into pBluescript KS ϩ . To generate a C␤1 fragment of 71 nucleotides with 5Ј-EcoRI and 3Ј-SmaI restriction sites, PCR primers 5Ј-GTTAACGAATTCGTCATCCC-TGCTTGCGGACTC-3Ј and 5Ј-CCTAGGGGTACCGCTCTCCACTTCG-CTGCCTTTCTT-3Ј were used. For production of a C␤2 fragment of 291 nucleotides with 5Ј-EcoRI and 3Ј-SacI sites, PCR primers 5Ј-GTTAAC-GAATTCGCTAGCTGGGGAAAAAAAAGC-3Ј and 5Ј-GAATTCGAGCT-CTACTCATGATGACAGGCAATG-3Ј were used. A 116-nucleotide fragment from C␤3 was made by PCR with 5Ј-EcoRI and 3Ј-SacI sites using primers 5Ј-GTTAACGAATTCCGAGGTCCGCAGCAGTAGGTG-3Ј and 5Ј-GAATTCGAGCTCACAAGCCCAAGTTTGATCCAG-3Ј. Antisense RNA probes were generated by linearizing at the EcoRI site and transcribing with T7 RNA polymerase. Probes were isolated on a Sephadex G-75 column, followed by ethanol precipitation.
Construction of C␤2 and C␤3 Expression Vectors-To produce expression vectors for C␤2 and C␤3, PCR was performed using a Zn 2ϩ -inducible C␤1 expression vector, C␤EV (15), to produce fragments with the cDNA sequence for C␤2 and C␤3. Primers were as follows: C␤2, 5Ј-G-AATTCCGGCCGCGCACGCCGCCATGAATGTGAAAGAGTTTCTAG-CCAAAGCC-3Ј; and C␤3, 5Ј-GAATTCCGGCCGCGCACGCCGCCGCC-ATGGGCTTGTTGAAAGAGTTTCTAGCCAAAGCC-3Ј. At the 3Ј-end of the sequence, a common primer was designed to introduce an ApaI site just following the termination codon of the C␤1 cDNA. This common primer was 5Ј-GAATTCGGGCCCGAGCTCCAGGTATCTCTCTCCTC-CCTA-3Ј. PCR products were cleaved with EagI and ApaI to release the coding fragment, which was ligated into EagI/ApaI-digested C␣ subunit expression vector to yield Zn 2ϩ -inducible, metallothionein promoterdriven expression vectors.
Cell Culture and Transfections-JEG-3 cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum as described previously (16). Cells were cultured in 24-well plates until ready for transfection (60 -70% confluent), at which time, the medium was reduced to 250 l. Four hours later, a mixture of calcium phosphate and DNA containing 2.5 ng of ␣-168-luciferase reporter construct, 50 ng of RSV-␤-galactosidase as a control for transfection efficiency, and the Zn 2ϩ -inducible metallothionein-C␤1, -C␤2, or -C␤3 expression vector was added. Expression vectors for RII␣ (ZR) (17), RI␣ (REV10) (18), and PKI (RSV-PKI) (19) were added in varying amounts as noted. After 24 h at 3% CO 2 , the medium was replaced by medium containing 2.5% fetal bovine serum and 80 M ZnSO 4 , and cells were returned to 10% CO 2 for 16 -18 h. Cells to be treated with forskolin were incubated with 30 M forskolin for 6 h. Forskolin stock (20 mM) was prepared in Me 2 SO, and Me 2 SO concentrations were held constant for all drug treatments. Induction of the luciferase reporter construct was normalized to the level of ␤-galactosidase activity used as a transfection efficiency control. Stable transfection of murine NIH 3T3 cells was performed using calcium phosphate precipitation and glycerol shock. The precipitate contained 100 ng of the selectable gene plasmid pKOneo (20) and 10 g of either of the Zn 2ϩ -inducible expression vectors metallothionein-C␤2 and -C␤3. Stable clones were selected and maintained in Dulbecco's modified Eagle's medium ϩ 10% fetal bovine serum and 500 g/ml Geneticin (Life Technologies, Inc.).
Primer Extension Analysis-Primers were designed to begin at the 3Ј-end of the first exons of C␤2 and C␤3. Sequences were as follows: C␤2, 5Ј-ATTCATGATGACAGGCAATGTCTC-3Ј; and C␤3, 5Ј-CAAGC-CCATGTTGGATCCAGCGAG-3Ј. Primers were end-labeled with 32 P and coprecipitated with 10 g of either mouse brain poly(A) ϩ RNA or yeast RNA. Samples were resuspended in hybridization buffer, denatured at 85°C for 10 min, and hybridized at 45°C for 8 -12 h. Hybrids were ethanol-precipitated, resuspended, and reverse-transcribed using SuperScript II reverse transcriptase (Life Technologies, Inc.). cDNA products were recovered, resuspended, electrophoresed on a 6% polyacrylamide denaturing gel, and analyzed by autoradiography. Sizing of primary transcripts was by comparison with M13mp18 sequence.
In Situ Hybridization-Coronal cryosections (20 m) from mouse brain were cut onto 3-aminopropyltriethoxysilane-treated slides. Slides were processed and hybridized as described previously (5). Briefly, slides with tissue were fixed in 4% paraformaldehyde, rinsed, treated with acetic anhydride, dehydrated through a series of ethanol rinses, and delipidated in chloroform. Processed slides were then hybridized with 35 S-labeled RNA probes specific for C␤1, C␤2, or C␤3. Controls were done with a 100-fold excess of the corresponding nonradioactive probe. Hybridizations were carried out overnight at 40°C (C␤1 and C␤3) or 55°C (C␤2). After hybridization, slides were washed at room temperature in 1 ϫ SSC (150 mM NaCl and 15 mM sodium citrate, pH 7.0). Slides were then treated with 20 g/ml ribonuclease (Promega) at 37°C for 30 min and washed three times (at 45°C for C␤1 and C␤3 and at 60°C for C␤2) in 0.1 ϫ SSC. Slides were dehydrated in a graded series of ethanol containing ammonium acetate and air-dried. Slides were apposed to Hyperfilm (Kodak) for 7 days. After the film was developed, slides were dipped in NTB2 liquid nuclear emulsion (Kodak), exposed for 21 days, developed, and counterstained with cresyl violet. supplemented with 1% fetal bovine serum. Cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum at 37°C in 10% CO 2 . Six h prior to induction of the transfected plasmids with ZnSO 4 , the cells were washed with Opti-MEM and fed with Opti-MEM ϩ 1% fetal bovine serum. After 18 h of treatment with 50 M ZnSO 4 , the medium was replaced with the [ 3 H]myristic acidcontaining medium. After 4 h, plates were washed twice with phosphate-buffered saline. Cells were solubilized with 1% Triton X-100 in a protease/inhibitor mixture. Cell extracts were subjected to SDS-polyacrylamide gel electrophoresis, fixed with 10% methanol and 10% acetic acid, treated with Amplify (Amersham Corp.), and dried under vacuum. The gel was exposed to X-Omat AR film overnight.

RESULTS
Genomic Structure of C␤-A mouse brain cDNA library was screened with a cDNA fragment corresponding to the 5Ј-end of the previously described C␤ cDNA (14). Positive clones were isolated and sequenced. Two novel C␤ cDNAs were found that diverged from the original C␤ sequence and had unique exon 1 sequences. Oligonucleotides corresponding to unique sequences within exons 1 of the C␤ variants were used to screen a panel of mouse genomic C␤ clones in Charon 4A that we had previously isolated (14). Two of these clones, 10 and 15, hybridized to one or more of these oligonucleotides. These clones were then mapped with a panel of restriction endonucleases and subjected to Southern blot analysis. In addition to the oligonucleotides specific for exons 1 of C␤1, C␤2, and C␤3, oligonucleotides were made to correspond to common sequence in exons 2 and 3. Blots of 10 and 15 were probed with these oligonucleotides, and exon 1 of C␤1 was located in 10. Exons 1 of C␤2 and C␤3 along with exons 2 and 3 were localized in 15 as shown in Fig. 1.
Labeled probes were made from EcoRI fragments of the two clones. Probes were then hybridized with Southern blots of the other clone. No cross-hybridization was detected, indicating that 10 and 15 do not overlap. This leaves a distance of at least 9 kilobases between C␤1 exon 1 and C␤3 exon 1. The 3.4-kilobase EcoRI fragment of 15 was sequenced, and the positions of the exons are indicated (Fig. 1). The restriction sites indicated are those mapped in 15 and confirmed when the fragment was sequenced.
Sequence Comparison of C␤ Splice Variants- Fig. 2A shows the sequence from the XbaI site to the end of C␤2 exon 1. Transcriptional (arrows) and translational (Met) start sites are indicated for C␤3 and C␤2. Sequences of the first exons of the three C␤ variants (Fig. 2B) are aligned at the splice junction of the common exon 2. C␤2 has only two amino acids in exon 1, and C␤3 has three amino acids compared with 15 amino acids in exon 1 of C␤1. C␤3 also has a T instead of a G in the first position of codon 4, which changes the first amino acid of exon 2 from valine to leucine. C␤1 and C␤3 have the N-terminal glycine residue that can provide a substrate for myristoylation, whereas C␤2 lacks this residue.
[ 3 H]Myristic Acid Labeling of C␤ Splice Variants-Mouse NIH 3T3 cells that overexpress each of the three C␤ splice variants were cultured as described under "Experimental Procedures." These cultures were exposed to 50 M Zn 2ϩ to induce the metallothionein promoter-driven expression vectors. Western analysis using anti-C subunit antiserum is shown in Fig.  3B. In the absence of Zn 2ϩ induction, the endogenous level of C␣ and C␤1 was observed as a single comigrating band. A previously described cell line overexpressing C␣ was also included as a positive control (11). In addition to induction of the expression vectors, Fig. 3B shows bands for C␤2 and C␤3 migrating more rapidly than those for C␤1, consistent with our predicted translational start sites and corresponding to the bands observed in brain extracts. Extracts from these cultures were run in parallel, and [ 3 H]myristic acid incorporation was detected using fluorography (Fig. 3A). Myristoylation occurred on C␣ and C␤1, but neither C␤2 nor C␤3 was a target for N-myristoylation Tissue-specific Expression Patterns of the Three C␤ Variants-Northern blot analysis of a panel of tissues isolated from both wild-type and C␤1 knockout mice revealed different patterns of expression for the three variants of C␤ (Fig. 4). C␤1 mRNA was found in all tissues examined, but the magnitude of expression varied, with the highest expression in the brain and a very faint signal in the testis. C␤2 had strong expression in the brain and was found in no other tissue examined. C␤3 was also found only in the brain, but at very low levels. It was necessary to isolate poly(A) ϩ RNA to detect a clear C␤3 signal. There was no compensatory increase in either C␤2 or C␤3 mRNA transcripts in brains from C␤1 mutant mice.
A peptide-specific antibody capable of recognizing all C␤ variants was used in Western blotting to demonstrate the unique expression of C␤2/3 variants in the brain (Fig. 4D). We could not distinguish between C␤2 and C␤3 because they comigrated on gels, but the relative abundance of C␤2 mRNA suggests that C␤2 protein accounts for nearly all of the lower molecular weight band expressed in the brain.
Functional Testing of C␤ Variants-Transient cotransfection assays were performed to determine whether the two novel C␤ variants produced active C subunit that can regulate expression of a cAMP response element-driven reporter construct. Both the C␤2 (Fig. 5B) and C␤3 (Fig. 5C) variants were capable of stimulating the cAMP response element-reporter expression to a level similar to that observed for the previously described C␤1 isoform (Fig. 5A). Each of the three variants was able to bind either type I or II R subunits as evidenced by inhibition of reporter gene transactivation. Activity was restored upon treatment with forskolin, which increases cAMP. Cotransfection with an expression vector for the heat-stable inhibitor of PKA (PKI) inhibited induction in a cAMP-independent manner for all C␤ variants. Thus, C␤2 and C␤3 behaved in this assay in a manner consistent with previous studies using C␣ or C␤1.
Many laboratories have relied on the use of specific chemical inhibitors of PKA to implicate this enzyme in regulatory pathways. A recent study demonstrated that the effects of cAMP on dopaminergic regulation of gene expression in developing striatum could not be blocked by the inhibitor H-89 (N-(2-((pbromocinnamyl)amino)ethyl)-5-isoquinolinesulfonamide HCl), and the authors suggested the possibility that brain-specific isoforms of PKA might exist that are insensitive to H-89 (21). Since C␤2 and C␤3 represent novel brain-specific catalytic subunits, we examined their H-89 sensitivity, and as shown in Fig. 5D, we conclude that these C␤ variants are inhibited effectively by H-89. We cannot rule out the possibility that other C␤ or C␣ variants exist that exhibit altered H-89 sensitivity.
In Situ Hybridization of Coronal Sections of Mouse Brain-Autoradiographic analysis of cryosections of mouse brain hybridized with C␤ isoform-specific riboprobes revealed very different distributions of the three mRNAs (Fig. 6). The C␤1 isoform was most highly expressed in the dentate gyrus and the pyramidal cell layers of the hippocampus. An intense signal was also observed in the habenula. A generalized distribution was seen throughout the neocortex. C␤1 mRNA was also present in the thalamic areas, although no specific nuclei were evident. The same was true for the hypothalamus, with some localization in the dorsal medial and ventral medial hypothalamic nuclei.
C␤2 showed a very distinct and interesting pattern of expression. In the rostral part of the brain, intense labeling was seen in the prelimbic cortex. A concentration of C␤2 was also seen in the insular cortex. C␤2 mRNA was localized to the lateral septum, with the intermediate nucleus showing the most intense labeling. The bed nucleus of the stria terminalis was also strongly labeled with the C␤2-specific probe. The signal in the more caudal parts of the forebrain also differed from that of C␤1. The intense hippocampal signal seen with the C␤1 probe was absent, and only the dentate gyrus expressed C␤2 mRNA. The neocortex in this medial area of the brain contained a substantial amount of both C␤1 and C␤2 mRNAs. The amygdala displayed a prominent C␤2 signal, with lateral, medial, basolateral, and cortical amygdaloid nuclei evident. There was an absence of C␤2 hybridization in the thalamic area and a diffuse labeling of the hypothalamus, except for an intense label in the ventral medial hypothalamic nucleus. The signal for C␤3 was very faint, and its distribution was similar to that of C␤1. Competition with a 100-fold excess of unlabeled probe indicated that the signals observed are highly specific. DISCUSSION We have characterized two novel variants of the murine C␤ gene of PKA that are specifically expressed in the brain. These variants (designated C␤2 and C␤3) arise by the use of alternate first exons with distinct transcriptional start sites. This diversity in the N-terminal exon of PKA genes has been observed in species other than mouse. In Aplysia, a neuronal form of PKA has been shown to have two alternate N termini that can combine with either of two internal cassettes to produce enzymes with different biochemical characteristics (22,23). PKA C subunits with extended amino termini have also been described in Drosophila (24) and Dictyostelium (25). A cDNA encoding an alternately spliced form of the C␤ subunit has been cloned from bovine heart. This bovine C␤ variant has an alternate first exon that is spliced at the same place as the murine variants (26). Because the murine C␤ splice is in the identical location as in the bovine C␤ variant, our genomic clones were probed with an oligonucleotide made against the bovine sequence. No hybridization of this bovine sequence was seen with our murine genomic clones. However, because our genomic clones do not overlap, the presence of a sequence like FIG. 4. Tissue-specific expression of C␤ variants. A, 10 g of poly (A) ϩ RNA for wild-type (ϩ/ϩ) or C␤1 mutant (Ϫ/Ϫ) tissues was loaded onto alternate lanes. Blots were hybridized with riboprobes specific for each of the three C␤ isoforms and exposed for 3 days. B, total RNA (10 g) prepared from various mouse tissues was loaded onto each lane. The blot was hybridized with a riboprobe specific for C␤2 and exposed for 7 days. C, total RNA (10 g) or poly(A) ϩ RNA (10 g), as indicated, was loaded onto each lane. The blot was hybridized with a riboprobe specific for C␤3 and exposed for 7 days. D, 40 g of protein from wild-type (ϩ/ϩ) or C␤1 mutant (Ϫ/Ϫ) tissues was separated by SDSpolyacrylamide gel electrophoresis. The membrane was probed with a C␤-specific antiserum. The standard (Std) lane was loaded with 20 g each of extracts from C␤1-and C␤2-overexpressing cell lines. Skeletal M, skeletal muscle. the bovine C␤ variant in the intervening space is possible.
Analysis of the genomic sequence in the region that contains the alternate first exons of C␤2 and C␤3 reveals differences between these splice variants and the other PKA C subunits described previously (1). The murine C␣ gene and the C␤1 isoform have multiple sites of transcriptional initiation and a GC-rich 5Ј-flanking region typical of many constitutively active genes (14). In contrast, C␤2 and C␤3 have a single major transcriptional start site and an AT-rich promoter. Also of note is the presence of short upstream open reading frames in the 5Ј-leaders of C␤2 and C␤3. This structure is characteristic of genes that exhibit translational control (27). Our in situ results demonstrate cell type-specific control at the transcriptional level, but the presence of these upstream open reading frames suggests possible regulation at a post-transcriptional level as well.
Our functional testing reveals that the proteins made from the C␤2 and C␤3 sequences are able to activate a cAMPresponsive reporter gene. They are inhibited by R subunits and regain activity upon elevation of cAMP levels. They are also inhibited when PKI is overexpressed or the PKA-selective chemical inhibitor H-89 is present. There have been suggestions in the literature that neural C subunit isoforms might exist that could be resistant to typical PKA inhibitors (21). However, the experiment shown in Fig. 5D demonstrates the ability of H-89 to effectively block activity from these novel C␤ variants.
N-Myristoylation of a protein has an absolute requirement for a glycine residue at the amino terminus. Other substrate specificity parameters have been established and rely on the first six residues of the protein (28). Myristoylation of the C␣ subunit has been established (8), but the function in vivo of this post-translational modification remains unclear (11,29). The existence of C␤1 as a myristoylprotein is not unexpected due to sequence similarity to C␣. C␤2 lacks the glycine residue at the amino terminus and is not myristoylated. C␤3 has the Nterminal glycine, but does not incorporate myristate. The inability of C␤3 to act as a substrate for N-myristoyltransferase possibly resides in amino acids 5 and 6 of this protein. The acidic residue at position 5 may be incompatible with nega-FIG. 6. In situ hybridization analysis of C␤ variants in mouse brain. Wild-type mouse brain cryosections (20 m) were hybridized to RNA probes specific for C␤1, C␤2, and C␤3. Hybridizations were carried out overnight at 55°C (C␤1 and C␤3) or 60°C (C␤2  tively charged residues in the enzymatic cleft of N-myristoyltransferase. In addition, the bulky aromatic residue at position 6 is likely a stearic hindrance. We found C␤2 and C␤3 mRNAs to be expressed only in the brain. C␤2 has a high level of expression, approximately equivalent to that of C␤1 in preparations of whole brain. C␤3 is expressed at a low level in total brain RNA, and it does not appear to be regionally localized by in situ hybridization. Neither C␤2 nor C␤3 mRNA appears to compensate for the loss of C␤1 mRNA in null mutants. This was assessed both by Northern blots prepared from whole brain and by in situ hybridization experiments on wild-type and C␤1 mutant brains (data not shown). We also observed no increase in the lower molecular weight C␤ band on Western blots from C␤1 mutant mice, indicating no compensation at the protein level (12). 2 The differential distribution of C␤1 and C␤2 mRNAs in discrete areas of the mouse brain is intriguing. The appearance of prominent expression of C␤1 in the hippocampus was expected and is consistent with the defects seen in late-phase long-term potentiation and long-term depression in the Schaffer collateral-CA1 synapse (12). Previous work (5) has shown that C␣ is expressed in these same regions, but is apparently unable to functionally compensate for C␤1. It is possible that C␣ may not be expressed in the proper cells or in the correct subcellular location. Our data reveal only weak expression of C␤2 in the hippocampus, and clearly, C␤2 cannot compensate functionally.
The neocortex is labeled throughout by the C␤2 probe, with the prelimbic cortex and insular cortex being particularly intense. This is of interest, as these brain regions have been implicated in the pathophysiology of schizophrenia and are rich in D1 receptors that act by elevating cAMP (30). The septal region has strong signals for C␤2 in the lateral septal nucleus and the bed nucleus of the stria terminalis. In the more caudal aspect of the forebrain region, we see a strong C␤2 signal in the ventral medial hypothalamic nucleus. This nucleus has been implicated in alterations in feeding behavior and aggression. The amygdala, like the hypothalamus, is an integrator of input from other limbic structures and is labeled with the C␤2 probe in the cortical, medial, basal, and basolateral nuclei. We can only speculate as to the function of C␤2 in these areas. The amygdala and its homologous rostral extension, the bed nucleus of the stria terminalis, have some of the heaviest innervation by dopaminergic neurons in the brain, where C␤2 could be activated via D1 dopamine receptor binding. There are numerous interconnections among these regions, and they are believed to be involved in learning and memory, emotion, and more primitive aspects of behavior such as flight from danger or pain avoidance. The functional significance of these variants awaits targeted disruption of C␤2 and C␤3 and their effects on neural function and behavior.