Protein stability and domain topology determine the transcriptional activity of the mammalian glial cells missing homolog, GCMb.

The glial cells missing (GCM) family of transcription factors consists of Drosophila GCM and the mammalian proteins GCMa and GCMb. They are expressed in a highly restricted manner during development and are known or assumed to be important regulators of developmental fate decisions. As the biochemical properties of GCMb have not been studied so far, we have undertaken a detailed structure-function analysis of the mouse GCMb (mGCMb) protein. DNA-binding specificity was very similar to that of other GCM proteins. Nevertheless, mGCMb was only a weak transcriptional activator in a number of different tissue culture systems. Interestingly, this was not due to an intrinsic absence of transactivation potential. In effect, we were able to identify two separate transactivation domains within mGCMb, one carboxyl-terminally adjacent to the DNA-binding domain and the second within the extreme carboxyl terminus. Activity of both transactivation domains was, however, modulated by an inhibitory region unique to mGCMb and located between the two transactivation domains. Furthermore, pulse-chase experiments proved that the mGCMb protein has a half-life approximately four times shorter than mGCMa. Introduction of the above mentioned inhibitory domain of mGCMb into mGCMa shortened the half-life of mGCMa to a value typical of mGCMb with a concomitant reduction in transactivation potential. Given the strong correlation between protein stability and transactivation potential, functional differences between the two mammalian GCM homologs are likely due to differences in stability with a single inhibitory region in mGCMb being involved in the reduction of both.

The glial cells missing (GCM) family of transcription factors consists of Drosophila GCM and the mammalian proteins GCMa and GCMb. They are expressed in a highly restricted manner during development and are known or assumed to be important regulators of developmental fate decisions. As the biochemical properties of GCMb have not been studied so far, we have undertaken a detailed structure-function analysis of the mouse GCMb (mGCMb) protein. DNA-binding specificity was very similar to that of other GCM proteins. Nevertheless, mGCMb was only a weak transcriptional activator in a number of different tissue culture systems. Interestingly, this was not due to an intrinsic absence of transactivation potential. In effect, we were able to identify two separate transactivation domains within mGCMb, one carboxyl-terminally adjacent to the DNAbinding domain and the second within the extreme carboxyl terminus. Activity of both transactivation domains was, however, modulated by an inhibitory region unique to mGCMb and located between the two transactivation domains. Furthermore, pulse-chase experiments proved that the mGCMb protein has a half-life approximately four times shorter than mGCMa. Introduction of the above mentioned inhibitory domain of mGCMb into mGCMa shortened the half-life of mGCMa to a value typical of mGCMb with a concomitant reduction in transactivation potential. Given the strong correlation between protein stability and transactivation potential, functional differences between the two mammalian GCM homologs are likely due to differences in stability with a single inhibitory region in mGCMb being involved in the reduction of both.
Many important transcriptional regulators of vertebrate development have been identified by their homology to Drosophila proteins. The functions of the vertebrate homologs are often similar to those of their Drosophila counterparts. While studying Drosophila mutants with nervous system defects, a gene was isolated that was transiently expressed in cells of the nervous system destined to become glia (1)(2)(3). Because flies in which this gene was inactivated showed a selective loss of all glial cells in the nervous system with the exception of midline glia, the gene was named glial cells missing (gcm). Detailed inspection of the phenotypes resulting from gene inactivation and transgenic overexpression revealed that GCM 1 is in fact the earliest marker known for cells of glial fate and is directly involved in the choice of an uncommitted precursor in favor of a glial and against a neuronal fate.
Biochemical analyses later showed that GCM had all the characteristics of a transcription factor with an amino-terminal DNA-binding domain and a carboxyl-terminal transactivation domain (4,5). At the time of its identification, GCM did not show any homology to proteins in the data bases. However, intense investigations have since led to the description of two mammalian proteins with sequence similarity to GCM (4 -9). These proteins are named GCMa (or GCM1) and GCMb (or GCM2) with a single letter prefix usually indicating the species of origin. Drosophila GCM and its two mammalian relatives thus make up a novel family of transcription factors.
Sequence homology between GCM and the mammalian GCMa and GCMb is restricted to the amino-terminal region, which contains the DNA-binding domain. Sequence comparison between the DNA-binding domains of Drosophila GCM and mouse GCMa (mGCMa) has in fact helped us to define the important structural determinants of the DNA-binding domain. Site-directed mutagenesis revealed the importance of seven symmetrically spaced cysteine residues and an adjacent lysine residue for integrity and binding capability of the DNAbinding domain (10).
The functions of the mammalian GCM family members are not well understood at the moment. Under the assumption that conservation of sequence is a good indicator for conservation of function, one might expect these proteins to have a role in mammalian gliogenesis. Developmental in situ hybridization studies, however, have so far only revealed GCMa expression in the embryonic trophectoderm (6,11,12) and GCMb expression in the forming parathyroid gland (9).
When mouse GCMa was expressed as a transgene in the developing nervous system of Drosophila, formation of surplus glial cells was observed (9,13). Ectopic expression in the fly thus seems to indicate that GCMa has very similar functional characteristics as Drosophila GCM. In agreement, both proteins behaved virtually interchangeably in transiently transfected cells when assayed for transcriptional activity (13). Similar to GCM, GCMa also had a strong transferable transactivation domain in its 80 carboxyl-terminal amino acids.
GCMb on the other hand behaved very differently when expressed as a transgene throughout the developing nervous system of Drosophila. Transformation of presumptive neurons into glia was not observed despite the fact that all parameters were identical to the ones used for ectopic GCMa expression (9).
Thus it was concluded that GCMb differs from GCMa and Drosophila GCM in its functional characteristics. To understand this difference in function and to obtain an idea of the structure of GCMb, we have undertaken a detailed comparative analysis of its properties in tissue culture.

MATERIALS AND METHODS
Plasmids-A fragment spanning the complete open reading frame of mGCMb (sequence similar to Ref. 9) was isolated from E15 mouse embryo cDNA and inserted into the mammalian pCMV5 and the insect pDS47/V5 expression vector. As previously reported for mGCMa (10), the amino-terminal 184 amino acids of mGCMb were amplified from this fragment by polymerase chain reaction and cloned between EcoRI and SalI sites of pCMV5 yielding the eukaryotic expression vector for the mGCMb DNA-binding domain. For detection, mGCMb constructs were also tagged with an amino-terminal T7 epitope (Novagen). Tagged and untagged versions behaved identically in all functional assays. To generate chimeras between mGCMa and mGCMb, fragments corresponding to the respective amino-terminal DNA-binding domains or all of the following sequences were generated by PCR and combined crosswise making use of an introduced SalI site before final insertion into pCMV5 (for schematic representation see Fig. 4A).
An expression plasmid for mGCMa with an amino-terminal T7 epitope was constructed in a manner analogous to mGCMb. Using a polymerase chain reaction-dependent approach, amino acids 300 -345 were deleted from GCMa under simultaneous introduction of a BglII site, producing mGCMa⌬. Consecutively, the same BglII site was used to introduce amino acids 266 -352 of mGCMb into mGCMa⌬, yielding mGCMa⌬b. Both mGCMa⌬ and mGCMa⌬b were inserted into pCMV5 for eukaryotic expression. All other mGCMa expression plasmids and the luciferase reporter plasmid 6xgbs luc were as described (4,13).
pCMV-TPOU contained coding sequences for the DNA-binding POU domain of the Tst-1/Oct6/SCIP protein (14). Fragments corresponding to various regions of mGCMb were generated by polymerase chain reaction with flanking EcoRI and BamHI sites (see Fig. 6). Insertion of these sequences led to an in frame fusion, with mGCMb following Tst-1/Oct6/SCIP sequences. The POU-responsive luciferase reporter (HSVoct luc) contained one copy of the HSV ICP0 octamer binding site in front of the rat prolactin minimal promoter (15). The same mGCMb fragments were also inserted into pCMVGal4 (4). Transactivation potential of the resulting fusions between regions of mGCMb and the DNA-binding domain of Gal4 was analyzed using the 3xUAS luc reporter (4).
Transfections and Luciferase Assays-COS and 293 cells were maintained in Dulbecco's modified Eagle's medium, U138 human glioblastoma cells in RPMI 1640 medium (Life Technologies, Inc.). Drosophila S2 Schneider cells were kept in Drosophila Expression System medium (Invitrogen). All media were supplemented with 10% fetal calf serum. COS cells were transfected by the DEAE-dextran technique using a concentration of 500 g/ml DEAE-dextran followed by chloroquine treatment (16). All other cells were transfected by the calcium phosphate technique (15).
For extract preparation and metabolic labeling experiments, cells were transfected with 10 g of CMV expression plasmid/100-mm plate. For luciferase assays, cells were transfected with 2 g of luciferase reporter plasmid and 2 g of CMV expression plasmid/60-mm plate unless stated otherwise. The total amount of plasmid was kept constant using empty CMV vector. Cells were harvested 48 h after transfection, and extracts were assayed for luciferase activity (17).
Extract Preparation, Western Blots, and DNA-Protein Binding Assays-Cell extracts were prepared and analyzed in Western blots using the ECL detection system as described (16). The following primary antibodies were used: a polyclonal rabbit antiserum directed against amino acids 169 -436 of mGCMa, a polyclonal rabbit antiserum directed against the POU domain of Tst-1/Oct6/SCIP, a purified mouse monoclonal antibody directed against the T7 tag epitope (Novagen), and a purified mouse monoclonal antibody directed against the Gal4 DNAbinding domain (CLONTECH). Primary antibodies were used at dilutions ranging from 1:3000 to 1:12,000 and reacted with a suitable horseradish peroxidase-coupled secondary antibody.
Metabolic Labeling and Immunoprecipitation-48 h after transfection, COS cells on 100-mm plates were repeatedly washed and starved for 1 h in cysteine/methionine-free Dulbecco's modified Eagle's medium containing 1% fetal calf serum. Pulse labeling was performed for 1 h by addition of 35 S-labeled cysteine and 35 S-labeled methionine at a specific activity of 110 Ci/ml. After repeated washes, cells were placed back into regular Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum and harvested after varying incubation periods in ice cold radioimmune precipitation buffer (10 mM Tris-HCl pH7.6, 150 mM NaCl, 0.2% sodium-deoxycholate, 0.1% SDS, 0.1% Nonidet P-40, 1 mM dithiothreitol, 10 g/ml aprotinin, 10 g/ml leupeptin). The resulting cell lysates were incubated under constant rotation at 4°C with antibodies that were already coupled to bead-immobilized protein G or protein A. For immunoprecipitation of epitope-tagged mGCMb-based proteins, the monoclonal T7-tag antibody was used. Proteins derived from mGCMa and Tst-1/Oct6/SCIP were precipitated with polyclonal antisera directed against the respective proteins. The precipitates were washed four times before analysis on SDS-10% polyacrylamide gels.

RESULTS
Comparison of the amino acid sequence of mGCMb with those of mGCMa and GCM reveals significant homology only in the amino-terminal DNA-binding domain. Within the DNAbinding domain of mGCMa, we have previously identified several amino acids, in particular seven symmetrically arranged cysteine residues and an amino-terminally adjacent lysine residue, which are essential for strength or specificity of DNA binding or influence the domain's conformation (10). As shown in Fig. 1, all of these amino acids are present and conserved within the amino-terminal region of mGCMb leading to the assumption that mGCMb might have DNA-binding characteristics similar to mGCMa.
The preferred recognition element for Drosophila GCM had been determined to be 5Ј-ATGCGGGT-3Ј (4, 5). We have previously used a set of oligonucleotide probes with this consensus or single site mutations thereof to analyze DNA binding of mGCMa in detail (10).
Here we used this set of probes to analyze DNA-binding characteristics of mGCMb in electrophoretic mobility shift assays. Whole cell extracts from COS cells transfected with an expression plasmid for the DNA-binding domain of mGCMb served as a protein source. As observed for mGCMa and Drosophila GCM, mGCMb exhibited the highest affinity for the consensus motif ( Fig. 2A). Single point mutations of this motif influenced binding of mGCMb very differently. Changes in positions 3, 6, and 7 had the most severe effects; changes in positions 1, 4, and 5 were less dramatic. Intermediate effects were observed for changes in positions 2 and 8. Binding of mGCMb to this panel of oligonucleotides thus very closely resembles the pattern previously observed for both mGCMa and Drosophila GCM (10), indicating that this second mammalian GCM protein has DNA-binding characteristics not very different from the other members of the family.
However, closer inspection of DNA-binding characteristics exhibited minor differences between mGCMa and mGCMb. Using constant amounts of protein and probes with very similar specific activities, mGCMb reproducibly exhibited a higher affinity for the probe with mutation at position 4 (m4) than for the probe with mutation at position 5 (m5). Complex strength for the m4 probe was approximately 25% of the strength observed for the wild-type probe compared with approximately 10% for m5. In parallel experiments, mGCMa showed compa-rable affinities for both m4 and m5 (approximately 25% of the wild-type probe) as previously observed (10). It is unclear at present what these minor differences are caused by.
As mGCMb binds to the GCM consensus element with high affinity, it should be possible to analyze the function of mGCMb as a transcription factor in transiently transfected cells using luciferase reporters under the control of GCM consensus elements (4). mGCMb was expressed from its CMV-driven effector plasmid and was detected in its epitope-tagged version in COS cell nuclear extracts by Western blot analysis (Fig. 3A). mGCMa was detected using an antiserum directed against amino acids 169 -436 of this protein (Fig. 3A) or in its epitopetagged version by the same monoclonal antibody used for mGCMb (Fig. 3F).
When transient transfections were performed in U138 cells (Fig. 3B), mGCMb caused a 10-fold induction of a luciferase reporter with six tandemly arranged binding sites. In parallel transfections, mGCMa elicited a robust 180-fold stimulation of the same reporter. Thus, mGCMb is a relatively poor transcriptional activator in U138 cells when compared with mGCMa.
To analyze whether this effect was cell line-specific or a general phenomenon, we carried out additional transfections with the same luciferase reporter in a number of other cell lines. When COS cells were used, we obtained a 3-fold induction for mGCMb versus a 10-fold induction for mGCMa (Fig.  3C). In 293 cells, activation rates for mGCMb were higher, approximately 20 -30-fold. Still, mGCMb was clearly less effective than mGCMa, which on average elicited a 100 -110-fold stimulation in parallel transfections (Fig. 3, D). Thus, there is significant variation in the transactivation capacity of mGCMb depending on the cell line used. These differences notwithstanding, mGCMb is always much less active than mGCMa.
As functional differences between both mammalian GCM homologs were originally observed in transgenic flies (9), we repeated our comparative transactivation studies in Drosophila S2 Schneider cells (Fig. 3E). Reporter gene activation achieved with mGCMb was again low. On average we obtained an 11-fold stimulation for mGCMb, compared with a 98-fold stimulation for mGCMa. Thus, mGCMb exhibits only 11% the activity of mGCMa in Drosophila cells. This low transcriptional activity could in fact explain the different behavior of mGCMa and mGCMb transgenes in Drosophila.
All transactivation studies reported so far were carried out under conditions of maximal reporter gene activation. Using increasing amounts of expression plasmids for the epitopetagged versions of mGCMa and mGCMb, we next analyzed reporter gene activation in relation to steady-state levels of both proteins in transfected 293 cells. For any given amount of expression plasmid, activation rates were higher for mGCMa than for mGCMb (Fig. 3F). Even with the highest amount of mGCMb expression plasmid, reporter gene induction remained significantly below the activation rates obtained with as little as 50 ng of mGCMa expression plasmid.
Western blot analysis with an antibody directed against the epitope tag was used to study the steady-state levels of GCM proteins in the transfected 293 cells. Even though the same expression system was used for both proteins, the amount of accumulated mGCMb protein was significantly below the amount detected for mGCMa for any given amount of expression plasmid (Fig. 3G). Thus, there is a good correlation between the reporter gene activation achieved by each GCM protein and its ability to accumulate in the transfected cell.
To determine the region responsible for the lower transcriptional activity of mGCMb as compared with mGCMa, we generated chimeras between both proteins. Chimera C1 contained the DNA-binding domain of mGCMa in the context of mGCMb sequences. Chimera C2 carried the DNA-binding domain of mGCMb with all other sequences being derived from mGCMa (Fig. 4A). Transient transfections were performed with these chimeras in 293 cells, as these cells not only allowed a clear distinction between mGCMa and mGCMb transactivation levels but also showed mGCMb transactivation levels significantly above background. In these cells, chimera C1 exhibited a transactivation capacity similar to mGCMb (29-fold for C1 and 32-fold for mGCMb), despite the fact that the aminoterminal domain was mGCMa-derived (Fig. 4B). Chimera C2, on the other hand, was even more active than mGCMa (180fold for C2 and 110-fold for mGCMa). Thus, transactivation of both chimeras does not correlate with the origin of the DNAbinding domain but instead with the origin of the sequences carboxyl-terminal to the DNA-binding domain. The latter sequences therefore have to be responsible for the functional difference between mGCMa and mGCMb.
In the case of mGCMa, the sequences following the DNAbinding domain contain two transactivation regions, one between amino acids 220 and 300 of mGCMa, the other from amino acid 349 to the carboxyl terminus (13). The presence of these transactivation domains is a good explanation for the strong transcriptional activity of mGCMa in transient transfection experiments. Both transactivation domains were transferable to the unrelated DNA-binding domain of the Gal4 transcription factor and functioned independently of each other.
To evaluate the presence of transactivation domains within those parts of mGCMb following its DNA-binding domain, we generated an in frame fusion of amino acids 174 to 504 to the Gal4 DNA-binding domain and assayed this protein for its ability to activate a reporter gene driven by a Gal4-dependent promoter (Fig. 5A). We failed to detect reporter gene activation independent of whether U138, COS or 293 cells were used (Fig.  5B, and data not shown). In parallel transfections, the comparable mGCMa construct (Gal4(aC) in Fig. 5B) showed the expected activation (13). However, when the expression of the Gal4/mGCMb fusion was checked by Western blot analysis with an antibody directed against the Gal4 DNA-binding domain, we likewise failed to detect protein of the correct size (Fig. 5C). Other fusions between the Gal4 DNA-binding domain and shorter regions of mGCMb also failed to be detected in Western blots, despite the fact that all constructions were repeatedly verified by DNA sequencing (Fig. 5C and data not shown). This argues that the Gal4/mGCMb fusions were highly unstable in the cells and that transactivation domains, if present, could not be detected by this method. We reasoned that the stability of mGCMb fusions might vary with the domain to which mGCMb regions are fused. Thus, we undertook a second attempt by fusing amino acids 174 to 504 to the POU domain of Tst-1/Oct6/SCIP (Fig. 6A). We first checked with an antibody directed against the POU domain of Tst-1/ Oct6/SCIP whether this fusion is at all expressed in transfected 293 cells (Fig. 6B). Western blot analysis revealed a protein of correct size indicating that this POU/mGCMb fusion is indeed Data are presented as fold inductions that were calculated for each reporter plasmid by comparing luciferase activities to values from cells transfected with reporter plasmid and empty expression plasmid. F, 6x gbs luc was transfected in 293 cells with increasing amounts of expression plasmids for epitope-tagged mGCMa or mGCMb (5 ng, 50 ng, 500 ng, 2 g, 4 g). The amounts of epitope-tagged mGCMa and mGCMb present in transfected 293 cells were determined by Western blot analysis using a monoclonal antibody directed against the T7 epitope (G). Equal amounts of extracts were used in each lane. All lanes were developed simultaneously and for the same length of time. The asterisks mark a cross-reactive band present in 293 extracts. more stable than the analogous Gal4/mGCMb fusion.
Using a luciferase reporter which is activated by POU proteins, we analyzed this mGCMb fusion for its transactivation capacity (Fig. 6C). The POU domain of Tst-1/Oct6/SCIP did not lead to significant transactivation of the luciferase reporter on its own and served as a negative control. The Tst-1/Oct6/SCIP holoprotein, on the other hand, elicited a 7-fold stimulation of reporter gene expression. Similar activation rates were also observed for the mGCMb fusion indicating that there is indeed a transactivation domain between amino acids 174 -504 of mGCMb.
To determine the borders of this domain, we generated further POU fusion proteins with mGCMb portions starting at amino acids successively closer to the protein's carboxyl terminus (Fig. 6A). The regions analyzed in these fusion proteins corresponded to amino acids 174 -504, 258 -504, 347-504, and 426 -504 of mGCMb. All were made at detectable levels in transiently transfected cells as evident from Western blot analysis (Fig. 6B). When tested for their ability to stimulate the POU-responsive promoter, the fusion with amino acids 258 -504 only yielded a 3-fold stimulation of reporter gene expression, which was significantly lower than the 8-fold stimulation obtained with amino acids 174 -504 (Fig. 6C). This result indicates that some of the transactivation potential is localized between amino acids 174 and 258, which in the context of mGCMb immediately follow the DNA-binding domain of the protein. Further shortening of the mGCMb portion again led to a significant increase in reporter gene expression, as evident from the 12-fold induction obtained with amino acids 347-504. Thus, the region around amino acids 258 -346 exerts a negative influence on the transactivation potential of mGCMb. With amino acids 426 -504 of mGCMb a 15-fold stimulation was obtained, indicating that the 79 carboxyl-terminal amino acids of mGCMb constitute a second independent transactivation domain. The localization of this transactivation domain correlates well with the localization of transactivation domains within mGCMa and Drosophila GCM (4,13).
To obtain an idea of the relative strength of this transactivation domain, we also fused amino acids 330 -436 of mGCMa to the POU domain of Tst-1/Oct6/SCIP. This region had previously been shown in the context of Gal4 fusions to contain the carboxyl-terminal transactivation domain of mGCMa (13). In  (13). B, the Gal4-responsive luciferase reporter (3x UAS luc) was transfected into 293 cells together with expression plasmids for Gal4 DBD or for the Gal4 DBD fusions depicted in A. Luciferase activities were determined in three independent experiments, each performed in duplicate. Data are presented for each Gal4 DBD fusion as fold induction above the level of luciferase activity obtained in transfections with an expression plasmid for Gal4 DBD alone, which was given an arbitrary value of 1. C, expression of Gal4 DBD fusion proteins was analyzed by Western blots of nuclear extracts from transfected 293 cells with a monoclonal antibody against the Gal4 DBD. Numbers indicate size of molecular mass markers in kDa. parallel transfections the mGCMa fusion elicited activation rates comparable to the ones obtained with the corresponding mGCMb fusion (amino acids 426 -504 in Fig. 6C).
A second deletion series of POU/mGCMb fusions was constructed with mGCMb fragments starting at amino acid 174 and extending variable distances toward the carboxyl terminus (Fig. 6A). A fusion that contained amino acids 174 to 353 of mGCMb exhibited a 13-fold induction of reporter gene expression compared with the 8-fold induction obtained with amino acids 174 -504. A fusion with amino acids 174 -263 of mGCMb yielded even higher induction rates as evident from the 29-fold stimulation in Fig. 6C. This result proves, on the one hand, that there are two completely nonoverlapping domains with transactivation potential in mGCMb. It confirms, on the other hand, the presence of an inhibitory domain contained within amino acids 258 -346 of mGCMb. Thus, mGCMb has a structure very similar to mGCMa with regard to its transactivation domains. The transactivation domains per se do not seem to be weaker than the ones found in mGCMa. The main difference between the two mammalian GCM proteins is the inhibitory domain in mGCMb, which therefore might be responsible for the significantly weaker transcriptional activity of mGCMb as compared with mGCMa.
Taking our observation into account that mGCMb did not accumulate to the same amount in transfected cells as mGCMa and that fusions between portions of mGCMb and the Gal4 DNA-binding domain were extremely labile, we reasoned that part of the functional difference between both mammalian GCM homologs could also be based on differences in protein stability. To compare the stability of mGCMb and mGCMa, we performed pulse-chase experiments in transiently transfected COS cells (Fig. 7). This allowed us to follow specifically the decay of protein synthesized de novo during a 1-h period in the presence of 35 S-labeled cysteine and 35 S-labeled methionine over a 6-h chase period. When quantifying the amount of radioactivity present within mGCMb immunoprecipitated from such cells, it became evident that approximately two thirds of the protein had already disappeared after a chase period of 40 min (Fig. 7, A and B). After 2 h, only 17% of the radioactivity was still detectable and 95% of all label had disappeared from mGCMb immunoprecipitates after 6 h. Accordingly, the halflife of mGCMb was estimated to be approximately 30 min in these cells. When this experiment was repeated under identical conditions with mGCMa, we still retrieved almost half the label in immunoprecipitates after a chase period of 2 h (Fig. 7C). Even after 6 h, 20% of the total label remained detectable. The half-life of mGCMa was therefore estimated to be close to 2 h and is therefore in the same range as the one observed for Tst-1/Oct6/SCIP in transiently transfected COS cells (Fig. 7D). Thus, we conclude that mGCMb has a significantly shorter half-life and significantly higher turnover than mGCMa.
We also determined the half-lives of C1 and C2, the two chimeras between mGCMa and mGCMb previously analyzed for their transactivation potential. In pulse-chase experiments, chimera C1, which had transactivation characteristics similar to mGCMb, had a half-life of approximately 30 min (Fig. 7E). Chimera C2, on the other hand, exhibited a half-life of approximately 3 h (Fig. 7F), thus resembling mGCMa not only in terms of transcriptional activity but also in terms of stability.
Both transcriptional activation capacity and stability of GCM proteins are determined by the region following the DNAbinding domain, suggesting a linkage between both properties. Considering the importance of the previously identified inhibitory domain for the overall transcriptional activity of mGCMb, it was of interest to analyze its impact on protein stability. We removed a region corresponding to amino acids 266 -352 from mGCMb and inserted it between the two transactivation domains of mGCMa (mGCMa⌬b in Fig. 8A). The mGCMa protein carrying the inhibitory domain of mGCMb was detectable by Western blot analysis, just as the mGCMa protein in which the spacer region between both transactivation domains had been removed prior to insertion of the inhibitory domain of mGCMb (mGCMa⌬ in Fig. 8B). When analyzed for their ability to activate a GCM-dependent luciferase reporter in transiently transfected 293 cells, mGCMa without the spacer region stimulated expression of the reporter as efficiently as the mGCMa holoprotein (Fig. 8C). Insertion of the inhibitory domain of mGCMb, however, led to a dramatic loss of reporter gene stimulation, reducing induction rates to levels comparable to those of mGCMb and corresponding to approximately 25% of those obtained with wild-type mGCMa. Insertion of amino acids 266 -352 of mGCMb also reduced the half-life of mGCMa as measured in pulse-chase experiments from approximately 2 h to less than 30 min (Fig. 8D). Thus, the inhibitory domain of mGCMb influences both protein turnover and transcriptional activity further supporting the close link between both properties. DISCUSSION The GCM family of proteins is a small, novel class of transcription factors consisting of Drosophila GCM and its two mammalian homologs mGCMa and mGCMb (5). Whereas GCM has a clear role in Drosophila gliogenesis (1)(2)(3), the function of its mammalian counterparts has yet to be defined. During embryogenesis, mGCMa is primarily expressed in trophoblasts of the placenta (11,12). mGCMb, on the other hand, was selectively detected in the parathyroid anlage (9), arguing that it might play a decisive role in the generation of this gland. The expression patterns therefore seem to argue against an evolutionary conservation of function within the GCM family. This conclusion is somewhat unexpected from studies on Drosophila GCM. Here, ectopic expression of GCM was not only sufficient within the nervous system to transform cells into glia but even outside it in mesodermal and epidermal tissues, arguing that GCM is not dramatically dependent on a particular cellular environment to exert its function (18,19).
Additionally arguing for a conservation of function, overexpression of mGCMa in the developing nervous system of Drosophila led to the same transformation of presumptive neurons into glia previously observed for GCM (9,13). In agreement, both proteins showed identical DNA-binding specificities in vitro and similar transactivation potential in tissue culture experiments (10,13). All amino acids within the DNA-binding domain, which proved to be relevant for mGCMa binding, were conserved and equally relevant for DNA binding of Drosophila GCM. In effect, a cysteine residue reported by us to be essential for DNA binding of mGCMa (10) was also found to be mutated to a serine residue in one of the fly mutants originally used to analyze the gcm phenotype (20). FIG. 7. Stability of mammalian GCM proteins. mGCMb, mGCMa, their chimeras, C1 and C2, as well as Tst-1/Oct6/SCIP were immunoprecipitated after various chase periods from pulse-labeled COS cells and analyzed on 10% SDS-polyacrylamide gels. A representative autoradiogram is shown in A for mGCMb. The amount of immunoprecipitated radioactivity was quantified using a phosphoimager with levels of radioactivity obtained for each protein without a chase period being arbitrarily set to 100%. The relative amounts were then plotted against the chase period for mGCMb (B), mGCMa (C), Tst-1/ Oct6/SCIP (D), chimera C1 (E), and chimera C2 (F).

FIG. 8. The inhibitory domain of mGCMb.
A, schematic representation of the two proteins mGCMa⌬ and mGCMa⌬b relative to mGCMa and mGCMb. Regions derived from mGCMa are shown as a filled box; regions derived from mGCMb are shown as an open box. B, expression and correct subcellular localization of mGCMa⌬ and mGCMa⌬b was confirmed by Western blot analysis of nuclear extracts from transfected COS cells. C, the GCM-responsive luciferase reporter (6xgbs luc) was transfected into 293 cells together with empty expression vector or with expression plasmids for mGCMa, mGCMa⌬, and mGCMa⌬b. Luciferase activities in extracts from transfected cells were determined in three independent experiments, each performed in duplicates. Fold inductions were determined as described, and data are presented as percent of the maximal activation rates obtained with mGCMa. D, mGCMa⌬b was immunoprecipitated after various chase periods from pulse-labeled COS cells and analyzed on 10% SDS-polyacrylamide gels. The amount of immunoprecipitated radioactivity was quantified using a phosphoimager with levels of radioactivity obtained for each protein without a chase period being arbitrarily set to 100%. The relative amounts were then plotted against the chase period.
Contrary to GCM and mGCMa, mGCMb has so far not been characterized biochemically. This study now fills the gap and provides several important insights into the biochemical properties of mGCMb. For one, we show that the DNA-binding characteristics of mGCMb are indeed very similar to those of the other GCM proteins. This is not only evident from the mobility shift experiments with variants of the GCM consensus binding site, but also from the strong transcriptional activation of promoters with GCM consensus binding sites by a chimeric protein that contained the DNA-binding domain of mGCMb. In agreement, all the amino acids previously identified as being important for mGCMa binding are fully conserved within mGCMb.
Despite this overall similarity, minor differences in the affinity toward GCM binding sites carrying particular mismatches were observed. This allows differential binding of mammalian GCM proteins to nonconsensus GCM binding sites and is therefore potentially relevant. The structural basis of these minor differences, however, cannot be determined without the availability of a crystal structure for a GCM-type DNAbinding domain and the possibility of comparative modeling.
Ectopic expression studies in transgenic flies had previously indicated that mGCMb was different from the other two GCM proteins because it did not cause the typical neuron-glia transformation (9). The failure of mGCMb to transform presumptive neurons into glia is unlikely to be due to differences in DNA binding. However, we reproducibly observed in a number of different cell lines that mGCMb was a significantly weaker transcriptional activator than either mGCMa or GCM. The amount of transcriptional activation obtained for mGCMb in different cell lines ranged from 5% to approximately 30% of the levels obtained with mGCMa. Importantly, mGCMb was only weakly active in Drosophila S2 Schneider cells, arguing that this difference in transactivation capacity might indeed be the cause for the behavior of mGCMb in transgenic flies.
Using chimeras between mGCMa and mGCMb, we were also able to show that the difference in transactivation potential is caused by those parts that follow the DNA-binding domain. These regions contain powerful transactivation domains, both in mGCMa and in Drosophila GCM (4,13). Thus it was reasonable to assume that the corresponding sequences of mGCMb might miss such strong transactivation domains.
However, this was not the case. Fusions between various regions of mGCMb and the transcriptionally inactive POU domain of Tst-1/Oct6/SCIP clearly showed that transactivation domains are present within mGCMb. In effect, these transactivation domains were found at similar positions to those previously mapped in mGCMa and GCM, thus further corroborat-ing the strong conservation of topology between members of the GCM family of transcription factors (Fig. 9).
Comparison of these transactivation domains with those of Tst-1/Oct6/SCIP and mGCMa showed that, at least in the context of POU domain fusions, they were not significantly weaker. Thus, we have to assume that the transactivation domains per se cannot be the underlying reason for the weak transcriptional activity of mGCMb. mGCMb was, however, unique in possessing a strong inhibitory domain located between amino acids 258 and 347 (Fig. 9). The activity of the adjacent transactivation domains was severely reduced in the presence of this domain. A region with comparable function had not been detected in mGCMa (13).
It was previously noted from sequence inspection that there are PEST sequences within mGCMb (7,8), which have the potential to mark the protein for rapid degradation (21,22). In total, there are three potential PEST sequences spread over wide parts of mGCMb that follow the DNA-binding domain (amino acids 204 -236, 270 -306, and 409 -437). mGCMa contains only one such sequence (amino acids 239 -261). Interestingly, the longest of the three PEST sequences in mGCMb is located within the region identified as a transcriptional inhibitory domain in this study (Fig. 9).
That protein stability could be an important determinant in explaining the functional differences between mGCMb and the other GCM proteins was indicated from two observations. First, mGCMb never accumulated to the same amount as mGCMa in transfected cells, and with equal amounts of expression plasmids there was always less mGCMb protein than mGCMa. Secondly, we failed to detect fusion proteins between the DNA-binding domain of Gal4 and mGCMb sequences. Why analogous fusions with the POU domain of Tst-1/Oct6/SCIP were detectable was not analyzed in detail. However, it seems reasonable to assume that the POU domain has a stabilizing influence on mGCMb regions.
Confirming the role of protein stability for transcriptional activity, we were able to show directly by pulse-chase experiments that in the same cellular environment mGCMb is much less stable than mGCMa. The half-life detected for mGCMb was approximately 30 min, whereas the half-life of mGCMa was close to 2 h and therefore not significantly different from that of the POU protein Tst-1/Oct6/SCIP. After 6 h, almost all of the mGCMb protein synthesized during the pulse period had disappeared, whereas 20 -30% were still present for mGCMa and Tst-1/Oct6/SCIP, respectively. Thus, mGCMb turns over with a half-life comparable to c-Myc which is known to be extremely labile (23). For comparison, many other transcription factors are significantly more stabile with reported half- FIG. 9. Structural comparison of GCM proteins. Domains were either defined experimentally as is the case for the DBD, the transactivation domains (TA1 and TA2), and the inhibitory domain, or were detected by sequence inspection in case of the PEST sequences (PEST). lifes ranging from 2 to 9 h (24 -26). Turnover characteristics of the mGCMb protein were determined by the region carboxyl terminally adjacent to its DNA-binding domain as evident from the half-lives of chimeras between mGCMa and mGCMb. A chimera with the DNA-binding domain of mGCMa and the carboxyl-terminal sequences of mGCMb had turnover and transactivation potential similar to mGCMb. Intriguingly, insertion of a region corresponding to amino acids 266 to 352 of mGCMb into mGCMa was sufficient to reduce both transcriptional activity and half-life of mGCMa to levels typical of mGCMb, indicating the importance of this region in determining the biochemical properties of mGCMb. As already mentioned, this region contains one of the three PEST sequences identified by sequence inspection in mGCMb. It is tempting to speculate that this region marks mGCMb for rapid degradation through its PEST sequence and thereby determines the low transcriptional activity of this protein. Functional differences between mGCMa and mGCMb thus are primarily due to differences in protein stability, and not so much due to different DNA-binding specificities or different strengths in the respective transactivation domains.