A potential dimerization region of dCAMTA is critical for termination of fly visual response

CAMTAs are a group of Ca 2+ /calmodulin-binding transcription activators that are implicated in brain tumor suppression, cardiac hypertrophy and plant sensory responses. The sole fly CAMTA, dCAMTA, stimulates expression of a F-box gene dFbxl4 to potentiate rhodopsin deactivation, which enables rapid termination of fly visual responses. Here we report that a dCAMTA fragment associated with a full-length protein in co-transfected HEK 293 cells. The interaction site was mapped to a region within the DNA-binding CG-1 domain. With this potential dimerization site mutated, the full-length dCAMTA had defective nuclear localization. In transgenic flies, this mutant dCAMTA variant failed to stimulate expression of dFbxl4 and rescue the slow-termination-of-light-response phenotype of a dCAMTA null mutant fly. Our data suggest that dCAMTA may function as a dimer during fly visual regulation, and that the CG-1 domain may mediate dimerization of CAMTA transcription factors.

A group of calmodulin-binding transcription activators (CAMTAs) have recently been identified in a variety of multicellular organisms. These include six in plants, two in mammals, one in flies, and one in worms (5,6). They all contain a DNA-binding domain (CG-1 domain) in the Nterminal, and two to four IQ motifs in the Cterminal region.
Using the fly CAMTA (dCAMTA) as an example, we have demonstrated that the CAMTA transcription activity is stimulated in vivo by direct interaction with calmodulin through the IQ motif region (7).
Compared to transcription factors such as the cAMP response element binding protein (CREB) and the nuclear factors of activated T cells (NFAT), which are stimulated through a series of calmodulin-dependent protein kinases or phosphatases (8)(9)(10), CAMTAs may respond to Ca 2+ /calmodulin in a more rapid and reliable manner.
A human CAMTA, CAMTA1, is expressed specifically in the brain, and has been identified as a candidate suppressor of both neuroblastomas and oligodendrogliomas (11)(12)(13)(14). The mouse CAMTA2 has been reported to mediate cardiac hypertrophy (15). Thus, it is of both theoretical and clinical significances to characterize the transcription activities of CAMTA.
Interestingly, the domain architecture of CAMTAs is very similar to that of the NFκB transcription factor precursor p100, p105 and Relish (16,17). Both groups of proteins have a TIG domain and several ankyrin repeats between the N-terminal DNA-binding domain and the C-terminal regulatory region. Considering that all NFκB transcription factors are either homo-or hetero-dimers (16)(17)(18)(19), it is important to examine whether CAMTAs may also function as dimers.
We have been using dCAMTA as a model to study the transcription activity of CAMTA, and have recently demonstrated that dCAMTA stimulates expression of a F-box and leucinerich repeat gene dFbxl4 both in the fly eye and in transfected HEK 293 cells (7). In fly photoreceptor cells, the dCAMTA-promoted dFbxl4 expression is required for rapid deactivation of the G protein-coupled light receptor rhodopsin. Due to the deficiency of dFbxl4, light responses in the eye of dCAMTA mutant flies terminate much slower than wild type (7).
In transfected 293 cells, dCAMTA stimulates expression of a luciferase reporter gene through a dFbxl4 promoter sequence that contains a CGCG box (7). Here we report that in the same system, a CG-1 domain-containing dCAMTA fragment (referred to as CG fragment) had much lower transcription activity than the full-length protein due to defective nuclear transport of the fragment. Remarkably, in cotransfected cells the CG fragment associated with and prevented nuclear transport of a fulllength dCAMTA protein. The interaction site was mapped to a region within the CG-1 domain. These data suggest that dCAMTA may dimerize through the mapped region. We further demonstrated that the potential dimerization of full-length dCAMTA protein is required for its nuclear localization and for its visual regulation function in the fly eye. (7).

Experimental Procedures
DNA construction and mutagenesis -The plasmid pCMV-dCAMTA-Myc has a full-length dCAMTA cDNA (including 21bp 5'UTR) inserted at the HindIII/XhoI sites of a pCMV-TAG5C Vector (Stratagene). Plasmids pCMV-FLAG-CG and pCMV-FLAG-TA contain the CG and the TA fragment, respectively, between the BamHI and EcoRI sites of pCMV-Tag2B (Stratagene). The plasmids for GST-CG and GST-TA fusions were constructed by inserting an EcoRV and a SalI/EagI fragment of dCAMTA into the SmaI and SalI/NotI sites, respectively, of the vector pGEX-5X-2 (Stratagene). Plasmids for truncated CG fragments were generated by restriction digestion and ligation of the pGEX-5X-CG construct. Site-directed mutagenesis of pGEX-5X-CG was conducted using mutationcontaining primers and a QuickChange kit (Stratagene). Mutant fragments were subcloned into a pcDNA3-dCAMTA plasmid to generate a full-length dCAMTA mutant, which was subsequently transferred into the vector pCMV-Tag2B or pCaSpeR-hs.
Cell transfection and immunohistochemistry -Using a Polyfect Transfection Reagent (Qiagen), the plasmids were introduced into 293 cells that were cultured on coated cover slips (2 cm, round) in 6 cm dishes. Two days after transfection, cells were fixed with 4% paraformaldehyde in PBS, washed twice, and incubated with mouse anti-Myc, rabbit anti-FLAG or both antibodies (diluted 1:200 in PBS, Sigma).
After washes, FITC and TRITCconjugated secondary antibodies (Sigma) were used to label the Myc and FLAG signals, respectively. The stained cover slips were mounted to slides using a DAPI-containing VectorShield medium (Vector Labs).
Luciferase assay -A dFbxl4 promoter sequence (-625~+87) was inserted upstream to a luciferase gene in a modified pGL3-basic vector (Promega), from which an intrinsic CGCG box (in the multiple cloning region) had been removed. The construct or the vector (for control) was introduced into 293 cells alone or together with the dCAMTA DNAs. After 24 hours, cells were harvested and examined with a luciferase assay system (Promega).
Glutathione-sepharose binding assay -GSTfusion proteins were expressed in BL-21 cells and purified with glutathione-sepharose beads (Amersham). Approximately 20 µg of each protein was coupled again to 40 µl Glutathionesepharose, and incubated with head extracts (in PBS that contains 1% Triton X-100 and protease inhibitors) of 200 heat shocked p[hs-dCAMTA];tes 2 flies (7), or with 0.5 ml 293 cell extracts in Tris-buffered saline (TBS). After three washes with TBS, proteins were eluted with 20 mM Glutathione in TBS and subjected to SDS-PAGE and Western-blotting.
Fly genetics -Both wild type and tes 2 flies had a w 1118 background. To generate dCAMTA transgenic flies, a wild type dCAMTA cDNA or mutant variants were subcloned into a pCaSpeRhs vector, and injected into w; tes 2 flies. All flies were raised in an approximate 12-hr light/12-hr dark cycle. To drive expression of dCAMTA through the heat-shock promoter, flies were shocked at late pupal stage by immersing the fly vials in 37°C water bath for 1 hour. All flies were examined at 1-2 days old.
Real-time RT-PCR -Total RNA was extracted from 1-3 day old fly heads using the Trizol reagent (Invitrogen). A 1-hour heat shock was applied to the fly 10 hours before the extraction. Real-time RT-PCR was conducted using an ABI PRISM7700 and a SuperScript III Platinum One-Step kit (Invitrogen). The raw mRNA level was determined as 2 -C T, where, C T is the number of cycles required for the SYBR green fluorescence to cross the threshold of 30 arbitrary fluorescence units. The relative mRNA levels were calculated after setting the raw level of each gene in the WT fly as 100%. The primer pair for dFbxl4 are cgatgctaacgaaaacgctg/ cggacgttgagccaatagc, which produce a PCR fragment of 189 bp.
Electroretinogram recording -ERG was examined as previously described (20). The fly eye was stimulated with 5-sec. orange light pulses (4000 lux). To quantitate the speed of response termination, the time required for 1/3 recovery was measured. For each genotype, data of 7 flies were averaged, and the standard error of mean (SEM) was calculated and presented as error bar.
Electrophoresis mobility shift assays -The DNA probe CCAACAGTCGCATGGGCAG CGTGCCCACGCG(or CGGG)CACCATTGG CGCCAGTATGAG was synthesized with overlapping primer pairs using PCR, and labelled with 3' biotin (Pierce). One micrograms of GST or GST-CG fragment (with R 483, 484 E mutations) were incubated with 10 fmol labelled probes, either with or without the presence of 2 pmol unlabelled probes, in 20 µl binding solution (Pierce lightshift chemiluminescent EMSA kit) for 20 min, and then loaded to 6% PAGE gel. The signal was developed according to the manual included in the kit.

Results
A CG-1 domain-containing fragment localized a full-length dCAMTA protein in the cytoplasm -In a study for mapping functional domains of dCAMTA, we found that a CG-1 domain-containing fragment (CG fragment, Fig.  1A) was able to stimulate expression of a luciferase reporter gene through a dFbxl4 promoter (7) in transfected HEK 293 cells (Fig.  1B). Thus, this fragment may also contain a transcription activation region that is next to the CG-1 domain (5,15). Nonetheless, the CG fragment displayed a much lower activity of transcription compared to a full-length dCAMTA protein (1.37±0.33 vs 2.93±0.57, Fig.  1B).
To attempt to explain why this fragment is less efficient in transcription activation, we used tag peptide antibodies to examine the distributions of the fragment (with a FLAG tag) and the full-length protein (with a Myc tag) in the transfected cells. The result showed that in contrast to the nuclear localization of the fulllength, most CG fragment signals were observed in the cytoplasm (Fig. 1C), suggesting that a nuclear localization sequence (NLS) is missing in the fragment. The low level of transcription activity of the CG fragment could have been contributed by the residual amount of proteins in the nuclei.
Surprisingly, when the full-length protein and the CG fragment were transfected together into the same cells, both proteins failed to enter the nuclei (Fig. 1D). As a consequence, the luciferase expression level in the co-transfected cells (0.77±0.21) was even lower than that of the CG fragment alone (Fig. 1B). In the cytoplasm, the full-length protein and the fragment appeared to be co-localized in the same areas (Fig. 1D), which could be due to dimerization of these two proteins and that the formed heterodimer was somehow unable to enter the nucleus. As the CG fragment displayed a higher protein level than the full-length in western blot (Fig. 1E), it is likely that most full-length proteins had been retained in the cytoplasm by the fragment.
A region in the CG-1 domain mediates the dCAMTA-dCAMTA interaction -If the CG fragment really dimerizes with the full-length dCAMTA protein, we should be able to detect a direct, biochemical interaction between these two. In a glutathione sepharose-binding assay, we found that the CG fragment, which was fused to a glutathione-S-transferase (GST), pulled down overexpressed full-length dCAMTA from head extracts of a transgenic fly (Fig. 2B). In addition to the CG-1 domain, dCAMTA also contains a TIG domain and an ankyrin-repeat region that may mediate protein-protein interaction. Nonetheless, a fragment (referred to as TA fragment, Fig. 2A) that contains both of these regions failed to bind any dCAMTA protein from the fly heads (Fig. 2B).
To examine whether the CG fragment interacts with its counterpart in the full-length protein, we attempted to pull down FLAGtagged dCAMTA fragments from 293 cell extracts using the GST-CG fusion protein. The result showed that the FLAG-CG, but not a FLAG-TA fragment, bound to the GST-CG protein (Fig. 2C). In control experiments, the GST-TA fusion protein did not pull down either fragment from the cell extracts (Fig. 2C).
The above observations suggest that a region in the CG fragment may mediate homodimerization of dCAMTA. To map this region, we truncated the GST-CG fusion protein into shorter fragments and tested their binding to the FLAG-CG. The fragments aa274-535 or aa274-509, but not the aa274-445, pulled down the FLAG-CG from cell extracts (Fig. 3A). The data indicate that the region aa445-509, which is entirely included in the CG-1 domain, is critical for the interaction between CG fragments. Most residues in this mapped region are conserved among CAMTA proteins from different species (Fig. 3B). We created two adjacent mutations, R 483 and R 484 to glutamic acid (E), in the GST-CG fragment. In glutathione sepharose-binding assays, these mutations completely disrupted the binding between the GST-and the FLAG-CG fragments (Fig. 3C), suggesting that the residues R 483 and R 484 are required for dimerization of the CG fragment.
To test whether the residues R 483 and R 484 mediate the CG fragment/full-length protein association in the 293 cells, we introduced a FLAG-tagged R 483, 484 E mutant CG fragment into the cell together with the full-length dCAMTA and examined their intracellular distributions. In the co-transfected cells, most fragment signals were still observed in the cytoplasm, while virtually all full-length proteins were localized in the nucleus (Fig. 3D). As the mutant fragment failed to prevent the nuclear transport of the full-length, the luciferase expression level in these cells was as high as that of the full-length dCAMTA alone (Fig. 3E). Thus, the residues R 483 and R 484 are indispensable for the interaction between the CG fragment and the full-length dCAMTA in live cells, and could be involved in a potential dimerization of dCAMTA.
The dCAMTA-dCAMTA interaction site is required for the transcription activity and the physiological function of dCAMTA in the fly eye -In the dCAMTA mutant flies tes 1 and tes 2 , the target gene dFbxl4 has much reduced expression level (7). To examine whether R 483 and R 484 are required for in vivo transcription activity of dCAMTA, we generated transgenic flies that express a R 483, 484 E mutant dCAMTA protein in the tes 2 background through a heat-shock promoter, and examined the expression level of dFbxl4. According to real-time RT-PCR, the dFbxl4 mRNA level in these flies was not significant increased compared to that in tes flies (Fig. 4). In contrast, overexpression of a wildtype dCAMTA protein completely recovered the level of dFbxl4 expression. The data indicate that R 483 and R 484 are critical for in vivo transcription activity of dCAMTA, probably due to their role in the potential dimerization of dCAMTA.
With insufficient dFbxl4, the light receptor rhodospin cannot be deactivated promptly in tes photoreceptors, and as a consequence the light response terminates much slower than wild type in both electroretinogram (ERG) and single-cell current recordings (7). This slow-terminationof-light-response phenotype was fully rescued by overexpression of wild-type dCAMTA proteins. Once the residues R 483 and R 484 had been mutated, however, dCAMTA failed to improve the termination speed of light response in the tes mutant background (Fig. 5). In contrast, mutation of two conserved residues (I 1852,1937 to N) in the C-terminal region did not significantly impair this visual function of dCAMTA (Fig. 5). Thus, the residues R 483 and R 484 are critical for the role of dCAMTA in rhodopsin regulation, which suggests that dCAMTA might need to form dimers to function in fly photoreceptor neurons.
The potential dimerization of dCAMTA is critical for its nuclear localization -To understand why the potential dimerization site is important for the transcriptional function of dCAMTA, we first examined the effect of R 483, 484 E mutations on the DNA binding of dCAMTA. In a electrophoresis mobility shift assay, the mutated CG fragment still bound to and changed the mobility of a biotin-labelled CGCG box-containing DNA probe (7) (Fig. 6A). This binding appeared to be specific because it was blocked by unlabelled probe of high concentration. In addition, changing the CGCG box into CGGG abolished this DNA-protein interaction (Fig. 6A).
As the dimerization did not appear to be critical for dCAMTA binding to specific promoter sequences, we next examined whether it is required for nuclear localization of dCAMTA by expressing a FLAG-tagged fulllength mutant protein in 293 cells. Immunostaining of the transfected cells showed that the majority of mutant proteins were localized in the cytoplasm and failed to enter the nuclei (Fig. 6B), suggesting that the R 483,484 -dependent dimerization could be important for appropriate presentation of the NLS. Thus, the mutant dCAMTA protein could be nonfunctional in the fly eye because of the impaired nuclear localization.
CAMTA transcription factors stimulate gene expression primarily through binding to CGCG box-containing promoter sequences (5,6). Here our data suggest that CAMTAs may act as dimers. Given that many transcription factors are assembled into both homo-and hetero-dimers (16,25,26), it is likely that CAMTAs could also hetero-dimerize with related transcription factors.
Because the heterodimers may recognize a promoter sequence different from that of homodimers (25,27,28), upon binding to Ca 2+ /calmodulin, a CAMTA heterodimer could even stimulate expression of genes that lack a CGCG-box promoter. Considering this, we may need to look for target genes of CAMTA in a much broader scope.
Similarly, dCAMTA appears to dimerize through a region within the CG-1 domain that is also involved in DNA binding. It has been reported that the CG-1 domain of mouse CAMTA2 mediates its interaction with another transcription factor Nkx2-5 (15).
Through this interaction, CAMTA2 functions as a co-activator of Nkx2-5 and enhances the Nkx2-5-mediated gene expression. Thus, the CG-1 domain may play a pivotal role in the assembly of CAMTA-involved transcriptional complexes.
In the future, it would be important to screen for additional CAMTA-interacting factors using a CG-1 domain-containing fragment.
The CG fragment failed to localize in the nucleus of the 293 cell, indicating that the CG-1 domain does not contain an effective NLS. A NLS has been mapped to the C-terminal region of mouse CAMTA2 (15). The observation that the R 483,484 E mutant dCAMTA is defective in nuclear localization may suggest that dimerization of the full-length protein is required for appropriate presentation of the NLS.
In co-transfected cells, the CG fragment may have competitively bound to the full-length dCAMTA and prevented dimerization between full-length molecules.
The cytoplasmic localization of the heterodimer suggests that dimerization with a CG fragment cannot expose the NLS in the full-length subunit. The fragment could even mask the NLS of the full-length protein. A similar case is that the P105 and IκB proteins bind to the NFκB subunit P65 and masks its NLS (16,30). Another explanation for the cytoplasmic localization of the heterodimer is that the fragment/full-length dCAMTA interaction may have exposed a strong nuclear export signal (NES) that is normally hidden inside the molecule. A NES has been identified in the N-terminal region of mouse CAMTA2, close to the CG-1 domain (15). No matter how the CG fragment may have retained the fulllength dCAMTA in the cytoplasm, a similar fragment could be used as a dominant inhibitor of mammalian CAMTA activities, and to prevent some CAMTA-mediated pathogenic processes such as cardiac hypertrophy.
In summary, our data suggest that dCAMTA may function as a dimer both in vitro and in fly photoreceptor neurons, and that the CG-1 domain may mediate the potential dimerization of CAMTA transcription factors.  promoter-mediated expression of luciferase in 293 cells that were transfected with a full-length dCAMTA, the CG fragment DNA or both. The co-transfected cells displayed the lowest luciferase activity. The pCMV vector DNA was used as control. Three sets of data were averaged and SEMs were shown as error bars. C, In 293 cells transfected separately with relevant DNAs, a full-length dCAMTA protein, but not the CG fragment, is localized in the nuclei. The proteins were stained with antibodies against the tag peptides, and the nucleus was labelled with DAPI. Dashed lines indicate the boundaries of cell. D, The full-length dCAMTA protein co-localized with the CG fragment and failed to be transported into the nucleus in the co-transfected cells. E, In co-transfected cells, the CG fragment had higher protein level than the full-length (f.l.). The western blot was probed with an antibody against the CG fragment (7). Untransfected cells were used as control.