Splice-specific Functions of Gephyrin in Molybdenum Cofactor Biosynthesis*

Gephyrin is a multifunctional protein involved in the clustering of inhibitory neuroreceptors. In addition, gephyrin catalyzes the last step in molybdenum cofactor (Moco) biosynthesis essential for the activities of Mo-dependent enzymes such as sulfite oxidase and xanthine oxidoreductase. Functional complexity and diversity of gephyrin is believed to be regulated by alternative splicing in a tissue-specific manner. Here, we investigated eight gephyrin variants with combinations of seven alternatively spliced exons located in the N-terminal G domain, the central domain, and the C-terminal E domain. Their activity in Moco synthesis was analyzed in vivo by reconstitution of gephyrin-deficient L929 cells, which were found to be defective in the G domain of gephyrin. Individual domain functions were assayed in addition and confirmed that variants containing either an additional C5 cassette or missing the C6 cassette are inactive in Moco synthesis. In contrast, different alterations within the central domain retained the Moco synthetic activity of gephyrin. The recombinant gephyrin G domain containing the C5 cassette forms dimers in solution, binds molybdopterin, but is unable to catalyze molybdopterin (MPT) adenylylation. Determination of Moco and MPT content in different tissues showed that besides liver and kidney, brain was capable of synthesizing Moco most efficiently. Subsequent analysis of cultured neurons and glia cells demonstrated glial Moco synthesis due to the expression of gephyrins containing the cassettes C2 and C6 with and without C3.1

ized to the large cytoplasmic loop of the GlyR ␤-subunit (3). Recently, GABA A receptor binding to gephyrin has also been reported, which involved the ␣2-subunit (4). GlyR binds with high affinity to the E domain of gephyrin (GephE) (5), and recent crystal structures have uncovered a key-and-lock mechanism of interaction between the receptor and GephE, which can be divided into four subdomains of which two participate in receptor binding (6,7). In addition, gephyrin interacts with a number of signaling, cytoskeletal, and vesicle-associated proteins that are believed to control gephyrin-mediated transport and clustering of GlyR and GABA A receptors (8). Based on crystal structures of individual gephyrin domains (7,9,10), the formation of a hexagonal gephyrin scaffold has been proposed (8). Interacting sites known from crystal structures were found to be important for gephyrin clustering (11), but alternative regions within the E domain also control gephyrin scaffold formation (12).
Besides the specialized function in neurons, gephyrin has a fundamental role in basic metabolism, catalyzing the final steps in molybdenum cofactor (Moco) biosynthesis (2,13), a cofactor of sulfite oxidase, aldehyde oxidase, and xanthine oxidoreductase (14). Moco contains a unique pyranopterin backbone with a terminal phosphate and a dithiolate group chelating the molybdenum (14). A deficiency of Moco results in early childhood death due to severe neuro-degeneration (15).
The biosynthesis of Moco can be divided into four major steps starting with GTP (14). The first two reactions generate the pterin backbone of the cofactor commonly referred to as molybdopterin (MPT). In animals, the last two steps, essential for molybdenum insertion, are catalyzed by gephyrin. The reaction mechanism has been uncovered for the gephyrin-homologous plant protein Cnx1, where the G domain binds MPT and catalyzes an adenylyl transfer to the terminal phosphate of MPT (16,17). Adenylylated MPT (MPT-AMP) is subsequently transferred to the E domain in a molybdate-dependent manner. Metal insertion is coupled to Zn-assisted hydrolysis of MPT-AMP (18) (Fig. 1A). Although in gephyrin the domains are inverted and connected by a much larger central domain (C domain), the same mechanism of Moco synthesis is expected for gephyrin (Fig. 1A), as it is able to reconstitute Cnx1 function (13). Finally, it is important to note that before molybdenum is inserted, a copper atom (Cu 1ϩ ) 4 is bound by MPT and released upon molybdenum transfer (16).
The gephyrin gene is ubiquitously expressed and shows high mosaicity. Eleven alternatively spliced exons were found in rat (19 -21), mouse (22), and men (23) (Fig. 1B). The nomenclature of these exons shows some irregularities because of different and independent identification. It is remarkable that except for two cases, all resulting mRNAs express full-length gephyrins without early termination. Gephyrin isoforms with the C2 and C6 cassette or an additional C4 exon in the C domain are the most common and first described variants (17) also referred to as P1 and P2 variants. A tissue-specific distribution of some splice variants (21,22) raised the question if the different functions of gephyrin in neurons as well as in basic metabolism may be undertaken by or restricted to different isoforms.
Here, we studied the Moco biosynthetic activity of eight gephyrin splice variants isolated from rat spinal cord (20) (Fig.  1C), by functional reconstitution of L929 cells, a gephyrin-de-ficient line for which we found a defect in the G domain of gephyrin (GephG). Individual domain functions were investigated in Escherichia coli mogA and moeA mutants. Both experiments demonstrated that C2 and C6 cassettes are essential for the gephyrin enzymatic function; cassettes in the C domain had no significant affect on the activity while the insertion of C5 resulted in a loss of GephG activity. Finally, we demonstrated that high levels of Moco are synthesized in brain, which was traced back to glial cells expressing metabolically active gephyrin variants.
vector were transiently expressed in L929 cells, grown as described above. Cells were seeded at a density of 5.4 ϫ 10 5 cells on 85-mm plates, transfected with EFFECTENE (Qiagen), harvested after 48 h, washed twice with phosphate-buffered saline, and sonicated in 100 -200 l of nit-1 buffer (50 mM sodium phosphate, 200 mM NaCl, 5 mM EDTA). For the determination of Moco and MPT, 3-30 l of crude extract were incubated with 20 l of Neurospora crassa nit-1 extract supplemented with 2 mM reduced glutathione. Moco and MPT were determined by reconstitution (overnight) of nit-1 nitrate reductase in the absence or presence of 5 mM sodium molybdate, respectively, as described (24).
Quantification of Expression Levels-Expression levels of the different gephyrin domain constructs and splice variants in L929 and E. coli cells were analyzed by Western blots using either mouse monoclonal anti-gephyrin antibodies mAb3B11 (1:25 dilutions) or polyclonal rabbit anti-gephyrin antibodies pAb-GepG (1:2000, "Puszta-Serum" (27)) as primary and anti-mouse or anti-rabbit alkaline phosphatase-coupled secondary antibodies (Sigma). Densitometric quantification of detected bands was performed with GeneTools software from SynGene, and the resulting differences were used to normalize the determined activities.
MPT Binding and Adenylylation Experiments-Gephyrin G domain as well as G domain containing the C5 cassette were expressed from pQE30-GephG (10) or pQE30-GephG5 (PCR cloned as pQE30-GephG) upon their transformation with pREP4 (Qiagen) into the E. coli strain BL21. Proteins were expressed and purified using a Ni-nitrilotriacetic acid column (Qiagen) as described (10) followed by size exclusion chromatography with a Superdex 200 XK16/60 column (GE Healthcare). MPT synthesis was performed in 100 mM Tris/HCl, pH 8.0 with 500 pmol of cyclic pyranopterin monophosphate (28), 100 pmol of in vitro assembled MPT synthase (29), and 200 pmol of GephG/GephG5 in a total volume of 140 l per reacting unit (time point) under anaerobic conditions. Adenylylation was induced with 0.5 mM ATP and 10 mM MgCl 2 , final concentration. The reaction was stopped at different time points by acidic iodine oxidation of MPT and MPT-AMP. Based on the formation of the fluorescent products FormA and FormA-AMP, the content of MPT and MPT-AMP was quantified by HPLC analysis as described (17).
Determination of MPT and Moco Content in Murine Organs-Organs (brain, heart, kidney, liver, lung) of 6-month-old mice were pestled in liquid nitrogen, solubilized in 100 mM Tris/HCl, pH 8.0, sonicated, and centrifuged for 20 min at 4°C. 100 and 300 g of crude extract were used to determine total MPT and Moco content by HPLC analysis as described above. The amount of Moco in all samples was further determined through the nit-1 reconstitution assay as described before.
Generation of Monoclonal Antibodies-Purified Geph-EE (5) was used to immunize mice following a standard immunization protocol. After hybridization and cloning, antibody producing hybridoma cells were screened for their binding to holo-gephyrin as well as the individual domains by ELISA. Supernatants were grown either according to standard protocols or in serum-free medium in the presence of 500 mg/liter albumin and 10 mg/liter transferrin. The resulting mAb3B11 was used as culture supernatant.
Cortex Cell Culture-Cortical cultures were prepared from E19 Wistar rats. All animals were sacrificed according to the permit (LaGeSo, 0122/07) given by the Office for Health Protection and Technical Safety of the regional government of Berlin and obeyed the rules laid down in the European Community Council Directive.
Harvested cell pellets were washed with phosphate-buffered saline buffer and sonicated in nit-1 buffer. Crude extracts were anaerobically concentrated with a 3-kDa concentrator (Amicon) to a protein concentration of approximately 10 mg/ml (with exception of the neuron-sample) and used for the nit-1 reconstitution assay.
While total MPT levels in L929 cells were comparable in all samples, including non-transfected cells (supplemental Fig.  S1A), clear differences were seen for the levels of reconstituted Moco ( Fig. 2A). Geph2,6 exhibited maximal activity, and no activity was found for GephE. Constructs with no or modified E domains showed a somewhat lower activity than full-length gephyrin, pointing to a functional advantage of the domain fusion. In conclusion, L929 cells express functional GephE, as isolated GephG is able to restore Moco biosynthesis.
Expression of transfected gephyrin variants was confirmed by Western blot using a newly generated monoclonal antigephyrin antibody (mAb3B11, for details see "Experimental Procedures") and a polyclonal anti-gephyrin antibody (pAb-GepG) that recognizes GephG (Fig. 2B) (27). As mAb3B11 only gave signals for constructs containing the entire E domain, we conclude that it recognizes an epitope located within GephE. It is important to note that no signal for endogenous gephyrin was detected with mAb3B11, although L929 complementation confirmed GephE activity. Successful expression of Geph-GC was demonstrated with pAb-GepG antibodies. Only Geph-GCE3/4 was hard to detect, probably due to a rapid degradation as reported recently (12).
These experiments were performed with EGFP-tagged proteins, similar to the following experiments with gephyrin splice variants in order to monitor transfection efficiency visually. To exclude any effect derived from the EGFP tag, we investigated the functionality of Geph2,6 as a representative control. Geph2,6 was expressed as non-tagged (pcDNA3-Geph2,6) or N-or C-terminally labeled EGFP fusion protein (pEGFP-Geph2,6). All constructs resulted in similar Moco synthetic activities in L929 cells (supplemental Fig. S2).
Consistent with the results from the domain complementation, the overall MPT content in cells expressing the different splice variants was comparable (supplemental Fig. S1B) and independent of reconstituted Moco levels (Fig. 3B). As seen for the domain reconstitutions, we found Moco synthetic activity for all gephyrin variants that contained G domains without splice modification. Variants with an additional C5 cassette (Geph2,5,6; Geph2,4,5,6; Geph1,2,4Ј,5,6) were completely impaired in Moco synthesis (Fig. 3B), suggesting a loss in GephG function. Although L929 cells were shown to be GephG-deficient, alteration at the E domain seems to impact the overall Moco synthetic activity of gephyrin, as deletion of the C6 cassette resulted in a significantly lower activity of Geph2. Interestingly, variants Geph2,6, Geph2,3,6, Geph2,4,6, and Geph2,4Ј,6 were in a similar range of activity, indicating that the insertion of cassettes C3, C4, and C4Ј into the C domain did not alter the overall function of gephyrin in Moco biosynthesis.
Equal expression of all eight gephyrin variants was confirmed by Western blot (Fig. 3C). Band intensities were densitometrically determined and used to normalize Moco levels to that of Geph2,6, resulting in even more similar activities (Fig. 3B, white  bars). Remarkably, for Geph2, no signal was detected with mAb3B11, while pAb-GepG antibodies showed a smaller but equally intense band corresponding to a 4-kDa reduction in molecular weight compared with Geph2,6, due to the missing C6 cassette in GephE. This finding suggests that mAb3B11 specifically recognizes residues within C6, which form the majority of subdomain 2 (see below). Therefore, mAb3B11 can be used as a splice-specific gephyrin antibody that binds only variants with the C6 cassette. Finally, the absence of any signal in Geph1,2,4Ј,5,6 confirmed the very weak expression of this variant (Fig. 3C).
Domain-specific Functions of Gephyrin Isoforms-Given the fact that L929 cells are defective in GephG, we could not investigate the functional properties of GephE in more detail. Therefore, individual domain functions of gephyrin variants were analyzed in E. coli by reconstituting Moco synthesis in the GephG-corresponding mogA mutant and GephE-corresponding moeA mutant, because bacteria express two separate proteins homologous to GephG and GephE, respectively. mogA mutants are unable to adenylylate MPT (16), while moeA mutants accumulate adenylylated MPT as they are defective in molybdenum insertion (18). Seven gephyrin splice variants were re-cloned into bacterial expression vectors (pQE30/ pQE80) and recombinantly expressed in both E. coli mutants. As gephyrin isoforms varied in their expression levels, different IPTG amounts were used for induction of protein synthesis. Crude extracts with comparable amounts of expressed gephyrin isoforms were identified by Western blots (Fig. 4, C and F), and Moco content was determined by nit-1 reconstitution (Fig.  4, A and D). Furthermore, the resulting Moco levels were normalized to the levels of Geph2,6 expression (Fig. 4, B and E).
Similar to L929 reconstitutions, all non-modified GephGcontaining gephyrin variants were able to restore Moco synthesis in mogA mutants, confirming the catalytic activity of their G domains (Fig. 4A). After normalizing the activities to the expression levels of each variant to the levels of Geph2,6 expression (Fig. 4, B and E), two forms, Geph2,3,6 and Geph 2,4Ј,6 were twice as active as the other positive gephyrin forms. Variants containing C5 were again inactive and indistinguishable from controls. Note that Moco levels were significantly higher than in L929 cells because of anaerobic growth conditions and stronger expression of gephyrins (expression was induced with 10 -200 M IPTG). Therefore, it cannot be excluded that MPT synthesis presents a limiting factor that might explain the increased normalized activity of variants with lower expression level (Geph2,3,6 and Geph2,4Ј,6). Furthermore, Geph2 shows almost the same activity as Geph2,6, while it was much less active in L929 cells. An explanation might be the low abundance of the remaining E domain, whose activity is lowered because of a dominant negative effect of the truncated E domain in Geph2. Such an effect is not seen in E. coli, because MoeA is strongly expressed.
In contrast to the L929 complementation, Geph2 also showed complete loss of activity upon expression in moeA mutants, because large parts of GephE are deleted in this splice form (Fig. 4D). All other variants contained high levels of Moco, which correlated with their expression levels (Fig. 4F), resulting in very similar activities (Fig. 4E). The overall Moco content in gephyrin-reconstituted moeA cells was 2-5 times higher than in mogA mutants, which goes back to the synthesis of eukaryotic Moco by gephyrin, while in MogA-deficient cells MoeA competes for MPT-AMP to drive the synthesis of the bacterial bis-MPT guanine dinucleotide cofactor (14) that is not detected in the nit-1 assay.
Biochemical Properties of C5-containing GephG-While the loss of activity in Geph2 was not surprising because of deletions of large parts of the active site within GephE (7), the loss of function in C5 cassette-containing variants was not so obvious. As only GephG function is altered in Geph2,5,6 and Geph2,4,5,6, we generated GephG with the C5 cassette (GephG5), expressed it in E. coli, and purified the protein to homogeneity as described previously for non-modified GephG (10). To further characterize GephG5, we performed size exclusion chromatography (Fig. 5A). In contrast to GephG, which is trimeric, GephG5 eluted at a molecular mass of 38 kDa, suggesting the formation of dimers as the theoretical mass of the GephG5 monomer is calculated to be 20.5 kDa.
To determine the in vitro activity of GephG and GephG5, MPT was synthesized from purified cyclic pyranopterin monophosphate (28), the first intermediate in the pathway, and in vitro assembled MPT synthase (29). Catalytic MPT synthesis is dependent on the presence of an MPT-binding protein, 4 which we use as a tool to measure substrate (MPT) binding to GephG5. While in the control experiment MPT synthesis was very low (2 pmol), both G domains, GephG and GephG5, were able to bind MPT (Fig. 5B). If the slope of MPT synthesis is taken as a measure of the ability to bind MPT, GephG5 shows a slightly reduced substrate binding.
Next, we investigated the catalytic function of GephG and GephGC, which has as yet only been studied for the plant homologue Cnx1G (17). While GephG showed a quantitative conversion of MPT into MPT-AMP, no activity was seen for GephG5, demonstrating a complete loss of function (Fig. 5C). These studies clearly demonstrate that insertion of C5 into the G domain of gephyrin alters its oligomerization without causing major changes in domain folding, because the ability to bind the substrate remained and only the catalytic function was abolished.

Distribution of MPT and Moco in Different Murine
Organs-Tissue/organ distribution of gephyrin splice variants has been reported previously using either mRNA expression (19,20,22,23) or proteomic analysis of GlyR-associated proteins (21), summarized in supplemental Table S1. Both studies confirmed that brain and liver are the organs with major gephyrin expression where most of the alternatively spliced cassettes were found. In addition, transcripts containing C2 and C6 cassettes were also identified in heart, kidney, and lung, but with lower abundance. Next, we investigated Moco and total MPT levels in these five tissues with proven gephyrin expression. Highest levels of MPT and Moco (measured together as the FormA oxidation product) were found in liver followed by kidney with the lowest readings in brain (3% of liver, Fig. 6A). However, the Moco content was found to be much higher in brain (22% of liver, Fig. 6B). This finding shows a more efficient conversion of MPT into Moco by gephyrins in neuronal tissue as illustrated by the Moco content with respect to total MPT, which is 7.3fold higher in brain than in liver (Fig. 6C). When comparing these data with gephyrin expression in the respective organs by Western blot using the mAb3B11 antibody, the strongest signals were found in brain followed by liver and kidney. The presence of at least three gephyrin bands in the brain extract points to several splice forms presumably generated by the splicing of cassettes located in the C domain.
Moco Synthesis in Glia Cells-To identify the cell type that contributes to Moco synthesis in brain, we prepared primary cultures of rat cortex neurons and glia cells that were subjected to Moco and gephyrin expression analyses (Fig. 7). Western blots using mAb3B11 showed that cortical neurons mainly express a gephyrin variant running around 95 kDa, while in glia cells an additional band, ϳ3-5-kDa larger than the lower one, was detectable (Fig. 7A). Transcript analysis by PCR indicated that cortical neurons mainly express gephyrin variants containing C2 and C6 with minor contribution of C4 and C5. The latter might be reflected by the weak signals seen above the major gephyrin band in Western blots (Fig. 7A). Sequence analysis of the C4-assigned PCR bands showed that three of four C4 cassettes identified in humans (23) are also present in a similar ratio in rat cortical neurons (C4a, C4c, and C4d, data not shown). In contrast, glia cells express mainly two gephyrin variants containing C2, C3, and C6 or only C2 and C6. Therefore, the larger gephyrin band seen in the glia Western blot can be assigned to C3-containing variants, which are similar in size to those variants detected in liver and kidney (Fig. 6D).
For Moco analysis, cells were harvested, extracted, and concentrated prior to use. Specific Moco activity was detected in glia cells, which was 30% higher than in total brain extract, indicating that glia cells mainly, if not exclusively, contribute to Moco synthesis in the brain (Fig. 7C). In contrast, no Moco activity was detectable in cortex neurons.
In summary, variants containing C2, C6 with and without C3 were found to be capable of synthesizing Moco in vivo. Moco synthesis in the brain is based on gephyrins expressed in glial cells, while gephyrins in neurons, biochemically able to synthesize Moco, do not seem to participate in Moco synthesis.

DISCUSSION
Gephyrin is a multifunctional protein that participates in at least two different processes: clustering of inhibitory neuroreceptors and biosynthesis of Moco. Alternative splicing and tissue-specific expression of gephyrin might contribute to the functional diversity including receptor binding (31), clustering (36), phosphorylation (37), binding to other proteins involved in gephyrin-mediated synapse formation, and/or controlling the gephyrin metabolic function in Moco synthesis. The latter was investigated in this study using eight splice variants that have been identified in rat spinal cord with modifications in both domains (G and E) essential for Moco synthesis as well as the C domain, which is believed to be crucial for brain-specific functions because it contains binding sites for a number of interacting proteins (8). Six of these variants represent the most abundant gephyrin forms identified in rat, mouse, and human and are found in different organs including brain and liver (supplemental Table S1).
First, we used gephyrin-deficient L929 cells (13) to study Moco synthesis of full-length gephyrin variants. Whereas Western blots with G and E domain-specific antibodies could not detect any gephyrin in L929 cells, we showed that only the G domain function is lost in those cells. Therefore, either very low levels of full-length gephyrin or an N-terminal-truncated form is expressed. The latter might be caused by mutations that result in the use of an alternate start site that has been proposed also for gephyrin variants lacking the C2 cassette (19). With respect to the functionality of gephyrin and its splice variants, it is important to note that all variants showed the same type of  aggregation seen in other non-neuronal cells (20,36) (Fig. 3A). Variants that form typical gephyrin "blobs" (Geph2,6, Geph2,3,6, Geph2,4,6, Geph2,4Ј6) were able to synthesize Moco with Geph2 as the only exception. Therefore, one can argue that these intracellular aggregates contain metabolically active gephyrin.
An important finding is that the C5 cassette inactivates GephG. C5 is inserted into ␣-helix 4 (10), which holds a loop in place that hosts a conserved Gly-Gly-Thr-Gly motif (Fig. 8A), in close proximity to the pyrophosphate bond in MPT-AMP (16). For the homologous plant Cnx1 G domain, Thr-542, located in the aforementioned loop, is vital for both MPT binding and catalysis (16,38). Furthermore, this loop also hosts an invariant aspartate residue, which is functionally important (39). Our in vitro studies on GephG containing C5 demonstrated that the overall folding of this domain is not altered, because GephG5 is able to bind MPT. However, catalysis of MPT-AMP synthesis, is completely destroyed by C5, which explains the loss of function in Geph2, 5,6 and Geph2,4,5,6. In addition, a significant change in the oligomerization shifting from trimeric GephG to dimeric GephG5 was observed. Previous gel filtration studies on C5-containing GephG were interpreted as a monomeric G domain because of an elution at 27 (36) and 21 kDa (11). Here, the observed elution at 38 kDa is closer to the theoretical mass of a GephG5 dimer (41 kDa). Therefore we favor a dimer, which could also explain the observed spike-like structures upon expression in non-neuronal cells (20) due to a uni-directional oligmerization of gephyrin dimers.
Recently, variants with the C5 cassette were mainly found in liver and to a somewhat lower extent in brain (21) (Fig. 7B). Previous findings have also shown that neuronal expression of C5-containing gephyrins might have a function in regulating inhibitory synapse composition (31). The presence of C5 in liver, as the main Moco-synthesizing organ, is somewhat surprising because C5-containing gephyrins were now found to be inactive in Moco synthesis. As C5 gephyrins show an altered oligomerization, they might help to disrupt gephyrin aggregates in order to maintain high levels of soluble gephyrin in liver. The fact that in brain Moco synthesis is most efficient suggests a very effective molybdenum insertion catalyzed by the gephyrin variants expressed in brain. The 7-fold higher Moco per MPT content in brain is partially mirrored by an increased expression of gephyrin (2-fold, Fig. 6D).
Two other cassettes in GephG, C1, and C2, also influence gephyrin stability/solubility and function in Moco synthesis (Fig. 8A). The C2 cassette was present in all examined splice variants, and previous data revealed that C2 is widely distributed in neuronal and non-neuronal tissue (21)(22)(23). A loss of C2 leads to an alternate start codon in the C-terminal region of GephG, resulting in N-terminal truncation by 115 residues (19) that would entirely destroy GephG function. Geph1,2,4Ј,5,6 harbors an additional C1 exon and showed very weak expression and/or rapid degradation with no detectable activity in Moco synthesis. The low abundance of Geph1,2,4Ј,5,6 might be related to structural changes at the N terminus of GephG because the individual insertions of C4Ј and C5 did not change protein expression/stability.
The only variant affecting GephE was Geph2, a form that was generated in vitro. The missing C6 cassette forms large parts of subdomain 2 (Fig. 8B), which forms the upper half of the GephE active site harboring residues important for Moco biosynthesis as shown for the homologous Cnx1 protein (40) as well as MoeA (41). The fact that Geph2 is expressed at similar levels as other variants (including blob formation) points to an overall maintenance of the structural integrity of GephE. It remains to be elucidated whether the binding of certain gephyrin ligands is dependent on the presence of C6 and might serve as an entry point to regulate gephyrin-clustering.
Recent progress in the understanding of the Moco biosynthetic function of the gephyrin-homologous Cnx1 protein (17,18) has uncovered the functional basis of domain fusion in Cnx1 as well as gephyrin (16 -18). Product substrate channeling was shown to be the main reason to fuse both domains within one protein. MPT-AMP, synthesized by the G domain, has to be efficiently transferred to the adjacent E domain. The main difference between Cnx1 and gephyrin is the inverted order of the domains, which might be one cause for the elongated C domain connecting the G and E domain in gephyrin. Consequently, this domain has gained unique features such as binding sites for proteins that control gephyrin-clustering and neuronal transport, including the GABA A receptor-associated protein GABARAP (42), dynein light chain proteins (43), or microtubules (22). Therefore it is not surprising that the C domain undergoes intensive alternative splicing.
L929 cells present a tool to investigate the function of fulllength gephyrins, and therefore our results further allow conclusions regarding the role of C domain-related splice variations. Geph2,6, Geph2,3,6, Geph2,4,6, and Geph2,4Ј6 showed similar activities in Moco synthesis, suggesting that splice cassettes in the C domain do not alter the function of gephyrin. Although previous studies reported a strong expression of Geph2,3,6 in non-neuronal tissue (23,44) such as liver (21) (Fig.  6), the organ with highest Moco synthesis required for sulfite oxidase catalytic activity (45), we could not detect increased Moco synthesis upon expression of Geph2,3,6.
As seen in Fig. 6D, the strongest expression of gephyrin is seen in brain with several splice forms detected by the E domain-specific mAb3G11 antibody. Consistent with our biochemical data, these splice variants should have Moco synthetic activity because high Moco levels were found with respect to the overall MPT content. Finally, we investigated if neurons, glia cells or both contributed to the observed Moco synthesis in brain. Our results clearly show that glia cells express two gephyrin variants (Geph2,6 and Geph2,3,6) at almost equal levels and that these variants are the main contributors to Moco synthesis in brain. To date, only residual activity of sulfite oxidase has been detected in brain. 4 Consequently, future studies need to investigate the expression and activity of Moco-dependent enzymes in brain tissue. Although, neuronal gephyrins were found to be able to synthesize Moco, no Moco synthesis has been detected in neurons. Therefore, we conclude that synaptic gephyrin does not participate in Moco synthesis suggesting that both the scaffolding and metabolic functions exist independent of each other. The latter is also supported by the finding that Moco-deficient MOCS1 knockout mice show wildtype-like gephyrin and GlyR clustering (46).
In summary, we have demonstrated that (i) the insertion of C5 into GephG abolishes Moco synthetic function of gephyrin, (ii) C6 is crucial for GephE function, and (iii) modification within the linker such as insertions of C3, C4, or C4Ј do neither affect individual domain functions nor the activity of the entire protein. These findings are in contrast to the GlyR binding properties of gephyrin splice variants (21,36) where none of the known splice cassettes are located in the GlyR-binding site of GephE (7) and therefore do not affect GlyR binding. (iv) Finally, we showed efficient Moco synthesis in the brain, which we traced back to glia cells, while cultured cortex neurons did not contribute to Moco synthesis. Therefore, we suggest that synaptic gephyrin is not involved in Moco metabolism.