Opposite Effects of KCTD Subunit Domains on GABAB Receptor-mediated Desensitization*

Background: GABAB receptors in the brain assemble with auxiliary KCTD8, 12, 12b, and 16 subunits that influence the receptor response in distinct ways. Results: Distinct KCTD domains exert opposing effects on desensitization of the receptor response. Conclusion: KCTD12 and -12b acquired desensitizing properties by disposing of a C-terminal inhibitory domain. Significance: This study defines the domains and motifs in KCTD proteins that generate functionally distinct GABAB receptor subtypes. GABAB receptors assemble from principle and auxiliary subunits. The principle subunits GABAB1 and GABAB2 form functional heteromeric GABAB(1,2) receptors that associate with homotetramers of auxiliary KCTD8, -12, -12b, or -16 (named after their K+ channel tetramerization domain) subunits. These auxiliary subunits constitute receptor subtypes with distinct functional properties. KCTD12 and -12b generate desensitizing receptor responses while KCTD8 and -16 generate largely non-desensitizing receptor responses. The structural elements of the KCTDs underlying these differences in desensitization are unknown. KCTDs are modular proteins comprising a T1 tetramerization domain, which binds to GABAB2, and a H1 homology domain. KCTD8 and -16 contain an additional C-terminal H2 homology domain that is not sequence-related to the H1 domains. No functions are known for the H1 and H2 domains. Here we addressed which domains and sequence motifs in KCTD proteins regulate desensitization of the receptor response. We found that the H1 domains in KCTD12 and -12b mediate desensitization through a particular sequence motif, T/NFLEQ, which is not present in the H1 domains of KCTD8 and -16. In addition, the H2 domains in KCTD8 and -16 inhibit desensitization when expressed C-terminal to the H1 domains but not when expressed as a separate protein in trans. Intriguingly, the inhibitory effect of the H2 domain is sequence-independent, suggesting that the H2 domain sterically hinders desensitization by the H1 domain. Evolutionary analysis supports that KCTD12 and -12b evolved desensitizing properties by liberating their H1 domains from antagonistic H2 domains and acquisition of the T/NFLEQ motif.

GABA B receptors assemble from principle and auxiliary subunits. The principle subunits GABA B1 and GABA B2 form functional heteromeric GABA B(1,2) receptors that associate with homotetramers of auxiliary KCTD8, -12, -12b, or -16 (named after their K ؉ channel tetramerization domain) subunits. These auxiliary subunits constitute receptor subtypes with distinct functional properties. KCTD12 and -12b generate desensitizing receptor responses while KCTD8 and -16 generate largely nondesensitizing receptor responses. The structural elements of the KCTDs underlying these differences in desensitization are unknown. KCTDs are modular proteins comprising a T1 tetramerization domain, which binds to GABA B2 , and a H1 homology domain. KCTD8 and -16 contain an additional C-terminal H2 homology domain that is not sequence-related to the H1 domains. No functions are known for the H1 and H2 domains. Here we addressed which domains and sequence motifs in KCTD proteins regulate desensitization of the receptor response. We found that the H1 domains in KCTD12 and -12b mediate desensitization through a particular sequence motif, T/NFLEQ, which is not present in the H1 domains of KCTD8 and -16. In addition, the H2 domains in KCTD8 and -16 inhibit desensitization when expressed C-terminal to the H1 domains but not when expressed as a separate protein in trans. Intriguingly, the inhibitory effect of the H2 domain is sequenceindependent, suggesting that the H2 domain sterically hinders desensitization by the H1 domain. Evolutionary analysis supports that KCTD12 and -12b evolved desensitizing properties by liberating their H1 domains from antagonistic H2 domains and acquisition of the T/NFLEQ motif.
GABA B receptors are the G-protein-coupled receptors (GPCRs) 5 for GABA, the main inhibitory neurotransmitter in the central nervous system. They are widely distributed throughout the brain and have been implicated in a variety of disorders including cognitive impairments, addiction, anxiety, depression, and epilepsy (1,2). GABA B receptors activate G␣ i/otype G-proteins that inhibit adenylyl cyclase and efficiently gate ion channels (3)(4)(5). Native GABA B receptors are known to comprise principal and auxiliary subunits that influence receptor properties in distinct ways (5,6). The principal subunits form two core receptors, GABA B(1a,2) and GABA B(1b,2) , that bind all GABA B ligands, couple to G-proteins and regulate classical GABA B receptor effectors, including G-protein coupled inwardly rectifying K ϩ channels (GIRK channels, also known as Kir3 channels) and voltage-gated Ca 2ϩ channels. The auxiliary subunits KCTD8, -12, -12b, and -16 are cytosolic proteins that modulate agonist potency and kinetic properties of the receptor response in distinct ways (6). In particular, KCTD12 or -12b produce fast desensitizing GABA B receptor-mediated Kir3 currents characterized by time constants of seconds while KCTD8 and -16 produce currents with little desensitization. The molecular determinants in the KCTD proteins and the mechanism underlying these kinetic differences in GABA B responses are unknown. Importantly, fast desensitization of GABA B receptor-mediated K ϩ currents is also observed with neurons (7,8) expressing KCTD12 (9). However, it has not been specifically addressed whether this fast desensitization is due to the presence of KCTD12 in the receptor. In addition, desensitization of GABA B responses was shown to be regulated by GRK4 (10), RGS proteins (11)(12)(13)(14) or phosphorylation of the GABA B2 subunit (15,16).
KCTD8, -12, -12b, and -16 constitute a subfamily of the KCTD family of proteins. All KCTD proteins contain a N-terminal bric-a-brac, tramtrak, and broad complex (BTB) domain that is most similar in sequence to the T1 tetramerization domain of voltage-gated K ϩ channels (17,18). In voltage-gated K ϩ channels, T1 domains are responsible for assembly of the four subunits around a central channel pore (19). Likewise, the T1 domains of auxiliary GABA B receptor subunits assemble into a homotetramer that tightly binds to the C-terminal intracellular domain of GABA B2 (6). All auxiliary GABA B receptor subunits contain a second conserved domain, designated the H1 homology domain, which is separated from the T1 domain by a non-conserved linker region. KCTD8 and -16 comprise an additional C-terminal conserved H2 homology domain. The H1 and H2 domains exhibit no sequence similarities to each other or to other proteins, thus giving no hints regarding their functions (5). Here we studied the influence of the H1 and H2 domains on desensitization of the receptor response. We found that H1 and H2 domains have opposite effects on the desensitization kinetics of the receptor response. The evolutionary analysis of protein sequences suggests that KCTD12 and -12b acquired desensitizing properties by disposing of their inhibitory H2 domains and selecting the T/NFLEQ motif in their H1 domains.

EXPERIMENTAL PROCEDURES
Generation of Expression Plasmids-All KCTD cDNAs were cloned in-frame with the N-terminal tag of three c-Myc epitopes (MEQKLISEEDLGEQKLISEEDLLEQKLISEEDLAAEF) into the cytomegalovirus-based expression vector pCI (Promega). Mutant constructs were generated using overlap extension PCR (20). To generate 12T1-16H1H2, residues Arg 207 to Glu 327 in KCTD12 were replaced by residues Arg 161 to Leu 427 of KCTD16. To generate 16T1-12H1, residues Arg 161 to Leu 427 in KCTD16 were replaced by residues Arg 207 to Glu 327 of KCTD12. To generate 12-16H2, residues Pro 280 to Leu 427 of KCTD16 were added in-frame at the C terminus of KCTD12. To generate 16H2, residues Pro 280 to Leu 427 of KCTD16 were cloned in-frame with the N-terminal c-Myc epitopes. To generate 16⌬H2, a stop codon was inserted after residue Glu 279 in KCTD16. To generate 12-16H2⌬60 and 12-16H2⌬113, a stop codon was inserted in 12-16H2 after residues Thr 367 and Cys 314 , respectively. To generate 12-Luc and 12-Venus, the cDNA of the Renilla reniformis Luciferase (Luc) or the GFP variant Venus from Aequorea victoria (Venus) was added inframe via a flexible peptide linker (GGGSGGGGS) to the C terminus of KCTD12. To generate Luc-12 and Venus-12, the cDNA of Luc and Venus, respectively, was added in-frame via a GGGSGGGGS peptide linker to the N terminus of KCTD12. N-terminally tagged KCTD12 constructs did not contain the three c-Myc epitope tag. Cell Culture-CHO-K1 cells stably expressing human GABA B1b and rat GABA B2 were maintained in Dulbecco's modified Eagle's medium (DMEM) with 500 M L-glutamine, 40 g/ml L-proline, 0.5 mg/ml G418, 0.25 mg/ml zeocine, and 10% FCS in a humidified atmosphere of 5% CO 2 at 37°C (21). Cells were transfected in 24-well plates at 80 -90% confluency using 3 l of Lipofectamine 2000 (Invitrogen) and 1.2 g of Kir3.1/3.2 concatamer in pcDNA3.1 (22), 2-g KCTD constructs in pCI and 0.3 g of pEGFP-N1 (Clontech) to visualize transfected cells. 6 h after transfection the cells were plated onto plastic coverslips (Thermanox, Nalge Nunc International) at a dilution of 1:5 in 35 mm dishes and used for electrophysiological recordings 24 -48 h later.
HEK293T cells (ATCC CRL-11268) were cultured in DMEM supplemented with 10% FCS and 500 M L-glutamine in a humidified atmosphere of 5% CO 2 at 37°C. Cells were transfected at 80 -90% confluency using Lipofectamine 2000. For transfection in 6-cm dishes, 12 l of Lipofectamine and 1.5 g of plasmid DNA were used. Cells were harvested after 48 h for co-immunoprecipitation and Western blot analysis.
Data were acquired with a MultiClamp 700B (Molecular Devices), low-pass filtered at 2 kHz and digitized at 10 kHz using a Digidata 1322A interface (Molecular Devices) driven by pClamp 10.2 software. Whole-cell currents were analyzed using Clampfit 10.2 software (Molecular Devices). All values are expressed as mean Ϯ S.D. Data were analyzed by one-way ANOVA followed by Dunnett's test for pairwise comparison with the control group (Igor Pro software). p values Ͻ 0.05 were considered as statistically significant.
Protein Alignments and Phylogenetic Trees-For the tree of all KCTD proteins, all annotated human KCTD protein sequences were used to query the NCBI nr protein database with blastp and all sequences with an e-value less than 1e-5 were retrieved and aligned using MUSCLE (23). Deeply diverging sequences that could not be reliably aligned were discarded (this set included all KCTD20 sequences). The sequences were re-aligned, and used as input into the tree-building program MrBayes 3.0 (24). The tree shown in Fig. 5A has been pruned to retain only the KCTDs from Mus musculus.
For the tree of KCTD8, -12, -12b, and -16, the H1 domain from human KCTD16 was used to query the NCBI nr protein database with blastp. The nucleotide sequences from all hits with an e-value less than 1e-10 were retrieved and aligned using MUSCLE. GBlocks (25) was used to retain only well-aligned blocks. This alignment was used as input into MrBayes. The tree shown in Fig. 5B has been pruned to retain only the KCTDs of M. musculus, Danio rerio, Branchiostoma floridae, Drosophila melanogaster, and Caenorhabditis elegans.
For the alignment of GABA B2 C termini, the human GABA B2 was used to query the NCBI nr protein database with blastp, and sequences from several species were aligned using MUSCLE. The C termini of non-vertebrates do not align with vertebrates and are not shown in the alignment.
The alignment of vertebrate KCTD H2 domain sequences was performed on the most recent ENSEMBL draft genomes of the species indicated. These sequences were blasted with the KCTD12b protein sequence from Oryzias latipes using tblastn. All hits which had H2-like sequences within the same contig were retained. All potential H2-like open reading frames were separated by 9.4 (Tetraodon nigroviridis) to 103 kb (Ornithorhyncus anatinus) from the open reading frame containing the T1 and H1 domains. The retained sequences grouped with KCTD12b and not KCTD8 or -16, suggesting that they are indeed KCTD12b sequences.  Fig. 1A. Previous work indicated that the lack of the H2 domain in KCTD12 and -12b correlates with strong desensitization of the receptor response (6). We therefore hypothesized that H1 domains facilitate and H2 domains inhibit desensitization. We tested this hypothesis using patch-clamp electrophysiology and CHO cells expressing GABA B(1b,2) receptors and effector Kir3 channels in the presence and absence of wildtype and mutant KCTD proteins. In the absence of KCTD proteins, activation of GABA B receptors for 40 s by GABA (0.1 mM) elicited Kir3 currents that slightly decreased in amplitude over time (Fig. 1A). The relative desensitization of GABA-activated Kir3 currents was calculated as the reduction in amplitude measured at the end of the GABA application normalized to the peak amplitude (Fig. 1C). In agreement with reported results (6), co-expression of KCTD12 significantly increased the desensitization of GABA-activated Kir3 currents, while KCTD16 had no significant effect on desensitization (Fig. 1, A and C; p Ͻ 0.001, compared with cells without KCTD). To identify the KCTD12 domain(s) responsible for desensitization we generated two chimeric proteins, 16T1-12H1 and 12T1-16H1H2, in which the H1 domain of KCTD12 and the H1/H2 domains of KCTD16 are swapped. The desensitization of GABA-activated Kir3 currents was significantly increased in cells expressing the chimeric protein 16T1-12H1 but not in cells expressing 12T1-16H1H2 (Fig. 1, B and C). Of note, both chimeric proteins contained the linker region of KCTD16. Western blot analysis confirmed that the chimeric KCTD proteins were expressed (Fig. 1D). In summary, these data show that the desensitizing properties of KCTD12 segregate with its H1 domain.

Distinct KCTD Protein Domains
We next tested whether the H2 domain exerts an inhibitory influence on desensitization. We generated the mutant 12-16H2 with the H2 domain of KCTD16 attached to the C terminus of KCTD12. 12-16H2 lacks desensitizing properties, in line with a dominant inhibitory effect of the H2 domain on KCTD12-mediated desensitization (Fig. 2, A and D). However, removal of the H2 domain from KCTD16 in the mutant 16⌬H2 does not produce a desensitizing KCTD protein (Fig. 2, A and  D). Therefore, the H1 domain of KCTD16 is not sufficient for desensitization and differs in its functional properties from the H1 domain of KCTD12.
We next addressed whether the H2 domain of KCTD16 not only prevents KCTD12-mediated desensitization in cis but also in trans. When the H2 domain of KCTD16 is co-expressed with KCTD12 as an independent 16H2 protein GABA B -activated Kir3 currents still desensitize (KCTD12 ϩ 16H2; Fig. 2, A and D; p Ͻ 0.001, compared with cells without KCTD). Expression of the 16H2 protein was confirmed by Western blot analysis (Fig. 2E). Therefore, the H2 domain only prevents KCTD12mediated desensitization in cis but not in trans. In order to map the minimal size of the H2 domain preventing desensitization in cis, we generated C-terminal truncations of the chimeric 12-16H2 protein. Truncation of 60 of the 148 amino acid residues of the H2 domain in the 12-16H2⌬60 protein is insufficient to restore desensitization (Fig. 2, B and D). However, truncation of 113 amino acid residues in the 12-16H2⌬113 protein fully restored desensitization (Fig. 2, B and D). This shows that the size of the H2 domain is critical for inhibition of KCTD12mediated desensitization. We tested whether adding KCTDunrelated protein domains to the C terminus of KCTD12 also prevents desensitization. Adding Luciferase (Luc) or the GFP variant Venus to the C terminus of KCTD12 in the 12-Luc and 12-Venus proteins completely prevented desensitization, suggestive of a sequence-unrelated steric hindrance of desensitization (Fig. 2, C and D). The 12-16H2, 12-Luc and 12-Venus proteins co-immunoprecipitate with GABA B2 (Fig. 2F). This demonstrates that the lack of desensitization of these proteins is not due to a lack of association with the receptor. Adding the Luc or Venus domains to the N terminus of KCTD12 in the Luc-12 and Venus-12 proteins did not prevent desensitization, showing that only protein domains C-terminal of the H1 domain are inhibitory (Fig. 2, C and D).
The Desensitization Motif of H1 Domains-We next determined the amino acid residues in the H1 domain of KCTD12 that differ from KCTD16 and mediate the desensitization. Sequence alignment revealed significant differences between KCTD12 and KCTD16 within the C-terminal half of their H1 domains (Fig. 3A). We therefore tested whether desensitizing properties can be transferred from KCTD12 to KCTD16 by replacing the C-terminal half of the H1 domain of KCTD16 with the corresponding sequence of KCTD12. We generated the two chimeric proteins 16T1-16/12H1G and 16T1-16/ 12H1N in which the 36 and 49 C-terminal residues, respectively, in the H1 domain of KCTD16 were replaced with those of KCTD12. In addition, we omitted in these chimeric proteins the H2 domain of KCTD16 to avoid its inhibitory effect on desensitization. The desensitization of GABA B receptor-activated Kir3 currents was small in CHO cell expressing 16T1-16/ 12H1G protein or the KCTD16. In contrast, cells expressing the 16T1-16/12H1N protein exhibited significantly more desensitization (Fig. 3, B and C; p Ͻ 0.001, compared with KCTD16). This result shows that the ability to desensitize GABA B -activated Kir3 currents can be transferred from the KCTD12 to KCTD16 H1 domain by exchanging the 49 C-terminal residues of the H1 domains. In addition, the result points at the 13 amino acid residues between Asn 277 and Ser 289 in the H1 domain of KCTD12 as being critical for desensitization. Sequence alignment of these 13 amino acid residues in KCTD8, 12, 12b, and 16 reveals that only Tyr 278 in KCTD8 is not identical or highly conserved with either KCTD12 or KCTD12b, which harbor an Phe residue at this position (Fig. 4A). We therefore addressed whether substitution of Tyr 278 in KCTD8 with Phe renders the H1 domain in KCTD8 desensitizing. We first established that deletion of the H2 domain of KCTD8 in the 8⌬H2 protein was insufficient to convert KCTD8 into a desensitizing subunit (Fig.  4, B and D), similar as already observed with KCTD16 (Fig. 2). Strikingly, substitution of Tyr 278 with Phe in the 8⌬H2F protein generated receptor responses with increased desensitization (Fig. 4, B and D; p Ͻ 0.05, compared with cells without KCTD). KCTD16 exhibits additional sequence divergence with KCTD12 and KCTD12b in the 13 amino acid residues under scrutiny (Fig. 4A). Accordingly, single amino acid substitution of His 232 , the residue homologous to Tyr 278 in KCTD8, with Phe in the 16⌬H2F protein was insufficient to render the H1 domain of KCTD16 desensitizing (Fig. 4, C and D). However, additional substitution of two neighboring non-conserved residues (Asn for Lys 231 and Gln for Arg 235 ) in 16⌬H2NFQ also rendered the H1 domain of KCTD16 desensitizing (Fig. 4, C and D; p Ͻ 0.01, compared with cells without KCTD). Finally, the converse substitution of these three residues in KCTD12 with the ones of KCTD16, but not the single amino acid substitution of Phe 278 by His, rendered the H1 domain of KCTD12 non-desensitizing (12KHR and 12H; Fig. 4, E and F; p Ͻ 0.001, compared with KCTD12). These experiments identify the motif T/NFLEQ in the H1 domain as a critical sequence element for KCTD-mediated desensitization.
Molecular Evolution of the KCTD Subunits-Our experiments show that the H1 domain is the functional unit respon-sible for desensitization of the receptor response. The H1 domains of KCTD8 and 16 lack desensitizing properties due to one or three amino acid substitutions, respectively, in the T/NFLEQ motif. The H2 domains in KCTD8 and -16 have antagonistic effects and inhibit desensitization by the H1 domains. To understand how the KCTD proteins acquired these regulatory domains we investigated their evolutionary history.
Analysis of the human and zebrafish KCTD proteins revealed that they are distinct from voltage-gated K ϩ channels, due to FIGURE 2. The H2 domain inhibits desensitization by the KCTD proteins. A-C, representative traces of GABA B -activated Kir3 currents recorded at Ϫ50 mV from CHO cells co-expressing GABA B(1b,2) , Kir3 channels, and KCTD proteins. In the presence of 12-16H2, a chimeric protein consisting of KCTD12 and the H2 domain of KCTD16, Kir3 currents exhibit modest desensitization. In contrast, when the H2 domain of KCTD16 was co-transfected with KCTD12 (KCTD12 ϩ 16H2), currents show significantly increased desensitization. This indicates that the H2 domain exerts an inhibitory influence on desensitization in cis but not in trans. However, KCTD16 lacking its H2 domain (16⌬H2) does not induce current desensitization (A). Deletion of 113 amino acids (12-16H2⌬113) but not of 60 amino acids (12-16H2⌬60) from the C terminus of the H2 domain in 12-16H2 restored the ability of the KCTD protein to induce current desensitization (B). Tagging KCTD12 with Luciferase at the C terminus (12-Luc) but not at the N terminus (Luc-12) eliminated its ability to induce desensitization (C). D, bar graph summarizing the desensitization of Kir3 currents in the absence and presence of KCTD proteins; wo KCTD, without KCTD. Data are expressed as mean Ϯ S.D.; *** , p Ͻ 0.001 compared with cells without KCTD (Dunnett's multiple comparison test). E, Western blot analysis of chimeric, truncated, or tagged KCTD proteins using anti-myc, anti-Luc or anti-GFP antibodies. The molecular mass is indicated on the left (in kDa). F, co-immunoprecipitation of C-terminally extended KCTD12 proteins with GABA B2 . The indicated Myc-tagged KCTD proteins were co-expressed with GABA B1b and GABA B2 (GB1 ϩ GB2). Immunoprecipitation was performed with antibodies against GB2 and immunoprecipitates were analyzed by Western blot with antibodies against GB1, GB2, and the Myc epitope. The asterisk likely indicates a truncated fragment of the 12-Venus protein that is not co-immunoprecipitated with GB2. Luc, Luciferase; Venus, Venus-GFP variant. differences in their T1 domains and the absence of transmembrane domains (17,26,27). Our phylogenetic analysis based on the amino acid alignment of the T1 domains of all annotated human KCTD proteins and their orthologues revealed that they diverged deeply in time, preceding the split of animals from plants. However, some KCTD proteins, including the subfamily formed by KCTD8, -12, -12b, and -16, diverged more recently (Fig. 5A). An ancestral KCTD protein with T1 and H1 domains having homology to this subfamily of KCTD proteins is found in nematodes, insects as well as invertebrate chordates ( Fig. 5B;  C. elegans, D. melanogaster, and B. floridae, a lancelet). However, the C-terminal GABA B2 domain mediating the interaction with the T1 domain (6, 28) is absent in invertebrate GABA B2 (Fig.  5C). Thus, it appears that the ancestral KCTD protein is not part of the invertebrate GABA B receptor complex.
Our phylogenetic analysis shows that soon after the emergence of vertebrates a number of events, occurring almost simultaneously in evolutionary terms, changed the structure of GABA B receptors. The C-terminal domain was added to the GABA B2 subunit, thus enabling interaction between the ancestral KCTD protein and GABA B receptors. Of note, the Tyr 902 residue in GABA B2 that is critical for binding to the KCTD proteins (6) is conserved in all vertebrates. In addition, the H2 domain was added to the ancestral KCTD protein. The ancestral KCTD protein then diversified into the KCTD8, 12, and 16 lineages (Fig. 5B). A suite of amino acid changes occurred in the H1 domain of two of these lineages: in the KCTD16 lineage, the TYLEQ motif changed to K/RHLER; in the KCTD12 lineage, the TYLEQ motif changed to NFLEQ (KCTD12) or TF/SLEQ (KCTD12b). In the KCTD8 lineage, the ancestral TYLEQ motif was kept. The final event in the evolution of this KCTD subfamily was a split of KCTD12 and -12b and the removal of the H2 domain in both sub-lineages. With the exception of placental mammals, most vertebrates retained a small part of the H2 domain as an open reading frame in their KCTD12b genes (Fig.  6). This demonstrates that the H2 domain was initially present  in the KCTD12 linage. It remains to be addressed whether the residual H2 domain sequences in the KCTD12b genes of vertebrates are transcribed and translated. In this respect, it is interesting to note that all H2 domain sequences analyzed, including those of KCTD8 and -16, are encoded by a separate exon downstream of the exon encoding the T1 and H1 domains. Therefore it is possible that multiple KCTD8, -16, or -12b variants are generated by alternative splicing. In conclusion, our phylogenetic analysis shows that desensitizing KCTD12 and -12b proteins evolved from non-desensitizing KCTD proteins by disposal of the inhibitory H2 domains and acquisition of the T/NFLEQ motif in their desensitizing H1 domains.

DISCUSSION
Association of auxiliary KCTD8, -12, -12b, and -16 subunits with principal GABA B receptor subunits was recently shown to generate molecularly and functionally distinct receptor subtypes (5,6,9,28). The four KCTD proteins are built from obligatory T1 and H1 domains and optional H2 domains. The T1 domains bind as tetramers to the GABA B2 subunit and thus are crucial for formation of the receptor complex. No functional roles have been assigned to the H1 and H2 domains of the KCTD subunits yet. In this study we show that the H1 and H2 domains have opposite effects on fast desensitization of the receptor response. H1 domains containing the T/NFLEQ motif mediate the desensitization while the H2 domains antagonize this desensitization. The antagonistic effect of the H2 domain is only observed when the domain is expressed in cis with the H1 domain but not when the H2 domain is expressed as a separate protein in trans together with KCTD12. This suggests that the H2 domain does not inhibit desensitization through the binding to a specific site. More likely, the H2 domain acts by sterically hindering the binding of the H1 domain to a downstream effector responsible for fast desensitization such as, for example, proteins involved in G-protein signaling. In agreement with a steric hindrance of desensitization by the H2 domain KCTDunrelated protein domains can substitute for the H2 domain and prevent KCTD12-mediated desensitization when tethered to the C terminus of KCTD12. It is unclear whether desensitizing and non-desensitizing KCTD proteins can simultaneously bind to the same receptor complex (6). If this is the case, our results suggest that the H2 domains of KCTD8 or -16 will be unable to inhibit desensitization by KCTD12 or -12b in trans. From an evolutionary perspective, it appears that the H2 domain was first acquired in an ancestral KCTD protein with non-desensitizing properties and subsequently lost in KCTD12 and -12b. It is therefore unlikely that the prime function of the H2 domain in KCTD8 and -16 is to antagonize desensitization.
KCTD12 and -12b evolved the T/NFLEQ motif within their H1 domain, which is necessary for KCTD-mediated desensitization of the receptor response. One to three amino acid substitutions within this motif can turn a desensitizing into a non-desensitizing H1 domain and vice versa. Secondary structure analysis predicts that the motif is part of a helix with amphipathic characteristics. It appears that an excess of positively charged amino acids within this helix, such as Arg, His, and Lys in KCTD16, is not permissive for desensitization. It is possible that these positively charged amino acids interact with negatively charged phospholipids of the plasma membrane and reduce the mobility of the helix. In addition, it appears that specific residues at the interface between the hydrophilic and the hydrophobic side of the helix are crucial for desensitization. Thus, substitution of Tyr 278 with Phe was sufficient to render the H1 domain of KCTD8 desensitizing. Interestingly, amphipathic helices are widely found in proteins participating in membrane-associated biological processes. In particular, amphipathic helices within GPCRs, G-protein subunits or naturally occurring peptides were shown to regulate the G-protein activation-deactivation cycle (29 -34). It is thus possible that the amphipathic helix in the H1 domain of KCTD12 and -12b directly regulates the G-protein that binds in its proximity to GABA B2 (35)(36)(37). However, no binding partners for the H1 domain have yet been identified.
Our evolutionary analysis shows that receptor subtypes owing to auxiliary KCTD subunits emerged with the appearance of vertebrates. In this respect it is interesting to note that GABA B1 subunit isoforms regulating axonal versus dendritic distribution of GABA B receptors (38) also first evolved in vertebrates. This suggests that it became essential to control GABA B receptor desensitization with the emergence of localized signaling. KCTD12 and -16 proteins appear to be present in pre-and postsynaptic GABA B receptors (6) albeit to differing degrees (9). Biochemical data support that certain GABA B receptors in the brain contain KCTD12 and others KCTD16 (6). However, whether association with specific KCTDs is responsible for the differences in desensitization between preand postsynaptic GABA B receptors remains to be addressed (39 -41). In conclusion, whereas the heteromeric nature and the activation mechanism of the GABA B core receptor are conserved in evolution (42)(43)(44)(45)(46), only the vertebrate GABA B receptors recruit functionally distinct auxiliary KCTD subunits and generate receptor subtypes.