Dual Role of the Molybdenum Cofactor Biosynthesis Protein MOCS3 in tRNA Thiolation and Molybdenum Cofactor Biosynthesis in Humans*

Background: E1-like proteins are required for activation and thiocarboxylation of β-grasp fold proteins involved in sulfur transfer to cofactors and tRNA. Results: MOCS3 interacts with both URM1 and MOCS2A in vivo and in vitro. Conclusion: Molybdenum cofactor biosynthesis and tRNA thiolation steps are linked by the MOCS3 protein in humans. Significance: The studies contribute to understanding the mechanism of protein conjugation and thiocarboxylate formation in sulfur transfer pathways. We studied two pathways that involve the transfer of persulfide sulfur in humans, molybdenum cofactor biosynthesis and tRNA thiolation. Investigations using human cells showed that the two-domain protein MOCS3 is shared between both pathways. MOCS3 has an N-terminal adenylation domain and a C-terminal rhodanese-like domain. We showed that MOCS3 activates both MOCS2A and URM1 by adenylation and a subsequent sulfur transfer step for the formation of the thiocarboxylate group at the C terminus of each protein. MOCS2A and URM1 are β-grasp fold proteins that contain a highly conserved C-terminal double glycine motif. The role of the terminal glycine of MOCS2A and URM1 was examined for the interaction and the cellular localization with MOCS3. Deletion of the C-terminal glycine of either MOCS2A or URM1 resulted in a loss of interaction with MOCS3. Enhanced cyan fluorescent protein and enhanced yellow fluorescent protein fusions of the proteins were constructed, and the fluorescence resonance energy transfer efficiency was determined by the decrease in the donor lifetime. The cellular localization results showed that extension of the C terminus with an additional glycine of MOCS2A and URM1 altered the localization of MOCS3 from the cytosol to the nucleus.

Ubiquitin (Ub) 3 and ubiquitin-like proteins (Ubls) are involved in a large number of diverse processes within the eukaryotic cell. The regulation is achieved by the covalent conjugation of the Ubls to target proteins via a lysine residue. Only recently, it was discovered that Ubls are not restricted to eukaryotes, with the identification of Pup in bacteria (1) and the characterization of SAMP proteins in archaea (2,3).
Ub-like protein conjugation is dependent on the activation of the C-terminal Gly-Gly motif in an ATP-dependent process, facilitating the formation of an acyl-adenylate and the subsequent thioester formation on a conserved cysteine residue of the activating enzyme (E1) and the Ubl. A cascade of enzymes (E2 and E3) is further involved in transesterification reactions for the transfer of the Ubl to a lysine residue of the target protein by formation of an isopeptide bond.
The E1-catalyzed activation of the Ubl resembles the second step of the molybdenum cofactor (Moco) biosynthesis in humans and bacteria. For Moco biosynthesis in humans, the E1-like protein MOCS3 forms a thiocarboxylate group at the C-terminal glycine of the ␤-grasp fold protein MOCS2A (4 -6). Moco is required for the activity of xanthine dehydrogenase, aldehyde oxidase, sulfite oxidase, and the mitochondrial amidoxime reducing components, hmARC1 and hmARC2, in humans (7). In Moco, two sulfur atoms of the molybdopterin (MPT) moiety coordinate the molybdenum atom in the final structure (8). Therefore, incorporation of two sulfur atoms in the first intermediate of Moco biosynthesis, cyclic pyranopterin monophosphate (cPMP), is required. The sulfur is mobilized from L-cysteine by NFS1, a pyridoxal phosphate-dependent L-cysteine desulfurase, which forms a persulfide group on its conserved Cys-381 residue (9). The persulfide group is further transferred to Cys-412 of the C-terminal rhodanese-like domain (RLD) of MOCS3 (10,11). The N-terminal MOCS3 domain shares amino acid homologies to ubiquitin-activating E1-like enzymes and to the Escherichia coli MoeB protein. In the activated MOCS3-MOCS2A acyl-adenylate complex, the sulfur from the C-terminal RLD is transferred to MOCS2A, forming a thiocarboxylate group at Gly-88 of MOCS2A and releasing MOCS3 and AMP (5). MOCS2A subsequently interacts with MOCS2B, which binds cPMP and generates MPT after the transfer of two sulfur atoms from two MOCS2A proteins (8). The major difference between the activation of Ub and the second step of the Moco biosynthesis is found in the underlying sulfur chemistry, because Ub activation involves thioester formation, and the thiocarboxylate formation is based on persulfide group transfer (12).
S. cerevisiae Urm1 was identified as the component of a new protein conjugation system by Furukawa et al. (18). Here, Urm1 conjugates to Ahp1p (alkyl hydroperoxide reductase), a protein involved in oxidative stress tolerance (19). Urm1 has some unique features because it has a dual function in protein conjugation and as a sulfur carrier (20); its structure resembles more those of sulfur carriers, like ThiS (21) and MoaD (22), which are involved in the biosynthesis of sulfur-containing cofactors, like Moco and thiamin, and therefore it presents a link between ubiquitin conjugation systems and the evolutionary older cofactor biosynthesis (23).
Only recently has it been shown that urmylation also occurs in humans and is involved in oxygen stress tolerance (23). It was shown that human URM1 is conjugated to lysine residues of target proteins in its own pathway and that oxidative stress enhances protein urmylation in mammalian cells. However, mechanisms underlying the dual function in protein conjugation and sulfur transfer are still controversially discussed, because on the one hand in cofactor biosynthesis persulfide formation is essential and on the other hand thioester formation is needed for Ub conjugation (12,23). Therefore, the mechanism of both conjugation and thiocarboxylation by Ubls has to be further elucidated. For such a purpose, URM1 displays an excellent example because it is combining both functions. As sulfur carrier, we proposed that MOCS3 connects the two distinct processes in the cell, for Moco biosynthesis and tRNA thiolation, especially because it is the only E1-like activating enzyme that is present in humans for these pathways (5).
In this study, we show that MOCS3 interacts with both URM1 and MOCS2A in human cells. We analyzed the cellular localization of the proteins in addition to their interaction by fluorescence resonance energy transfer (FRET). The FRET efficiency was calculated by determination of the decrease in the donor lifetime. Because the Gly-Gly motif is the one common sequence feature of the Ubls involved in both acyl adenylation and thioester or acyl-disulfide formation, the role of the terminal glycine was examined in MOCS2A and URM1, and the effects of the variations on both pathways were compared. In addition, we examined the importance of the C-terminal Gly-Gly motif in terms of its role for the localization of the proteins.

EXPERIMENTAL PROCEDURES
Bacterial Strains, Media, and Growth Conditions-Cell strains containing expression plasmids were grown aerobically at 30°C in LB medium containing 150 g/ml ampicillin or 50 g/ml chloramphenicol.
E. coli MoaE was expressed in BL21(DE3) cells from plasmid pMWaE15 and purified as described previously by Wuebbens and Rajagopalan (24). Human MOCS2A and URM1 wild type proteins were expressed and purified as described previously by Schmitz et al. (5).
Site-directed Mutagenesis, Protein Expression, and Purification-Single amino acid substitutions at the C termini of the human URM1 or MOCS2A proteins were created using PCR mutagenesis, resulting in URM1-G101A, URM1-G101⌬, and URM1-102Gϩ or MOCS2A-G88A, MOCS2A-G88⌬, and MOCS2A-89Gϩ variants, respectively. For expression of the variants, the URM1 cDNA was cloned into the pACYC-duet 1 BglII restriction site, and the MOCS2A cDNA was cloned into the sites NdeI and KpnI of pTYB2. E. coli BL21(DE3) cells and ER2566 were used for heterologous expression of URM1 and MOCS2A variants, respectively, as described previously (5).
Expression of MOCS3-Sf9 insect cells, derived from Spodoptera frugiperda (Invitrogen), were grown as suspension culture in Sf-900-II SFM medium (Invitrogen) supplemented with 2% fetal bovine serum (FBS, PAN-Biotech, Germany) at 28°C and 180 rpm. For infection with recombinant baculovirus, cells were grown as monolayer cultured at 28°C in a humidified incubator.
The restriction sites BamHI and XbaI were introduced into the MOCS3 cDNA by PCR, which allowed cloning into the pFastBacI vector (Bac-to-Bac baculovirus expression system, Invitrogen). The resulting plasmid was designated pMMC30 and expressed MOCS3 as an N-terminal fusion to a His 6 tag. Positive recombinant bacmids were used to transfect Sf9 cells using Cellfectin II (Invitrogen) and to produce a baculovirus stock. For protein expression Sf9 cells were infected with high titer viral stock (P2). Cells were harvested by centrifugation 72 h after infection and lysed by sonification, and the supernatant was subjected to Ni-NTA chromatography. MOCS3-RLD was expressed and purified as described previously (5,10).
Cell Culture Techniques-HeLa and HEK 293 cells were cultured in Dulbecco's modified Eagle's medium (DMEM, PAN-Biotech, Germany) supplemented with 10% FBS (PAN-Biotech, Germany). Cell cultures were maintained at a temperature of 37°C and 5% CO 2 atmosphere.
Construction of ECFP and EYFP Fusion Proteins-Enhanced cyan fluorescence protein (ECFP) and enhanced yellow fluorescence protein (EYFP) are among the most used FRET pairs (25). Because the donor fluorescence of ECFP is comparably dim, and sensitized emission of the acceptor is hard to detect due to the spectral bleed through of ECFP fluorescence, therefore the fluorescence lifetime of the donor was chosen as parameter to monitor FRET.
MOCS3 cDNA was amplified by PCR and cloned into pECFP-C1 (Clontech) vector using BamHI and SalI restriction sites, resulting in plasmid pZM13. MOCS2A and URM1 variants were cloned using SacI and HindIII restriction sites and ligated as N-terminal fusion proteins into pEYFP-C1 (Clontech) resulting in pMMC2 and pMMC4, respectively.
MPT Synthase Reactions-MPT synthase reactions were performed at room temperature in a total volume of 400 l of 100 mM Tris-HCl (pH 7.2). The produced MPT was converted to the stable oxidation product Form A and quantified as described by the published procedures (26,27). The reaction mixtures contained 5 M of E. coli MoaE, 5 M MOCS3, 2.5 mM Mg-ATP, 0.6 mM Na 2 S or sodium thiosulfate, and 15 M of each MOCS2A variant. The reaction was initiated by the addition of 2.5 M cPMP, which was purified as described previously (28).
Thiosulfate Sulfurtransferase Activity of MOCS3-The sulfurtransferase activity of MOCS3 was measured according to the procedure described by Sörbo (29) in 100 mM Hepes (pH 8.0). For determination of the K m and k cat values, each reaction contained 130 nM MOCS3, 5-500 mM sodium thiosulfate, and 0.1-50 mM potassium cyanide in a total volume of 500 l. Thiocyanate was quantified as iron complex by its absorption at 460 nm (⑀ ϭ 4200 M Ϫ1 cm Ϫ1 ).
MOCS3-catalyzed Adenylation of MOCS2A and URM1 Variants-For adenylation of MOCS2A and URM1 variants by MOCS3, 20 M of each protein pair was incubated at 25°C with 250 M Mg-ATP, and 2 units of inorganic pyrophosphatase in a total volume of 300 l. After 90 min, the reaction was stopped by addition of 1% SDS and additional heat inactivation at 95°C for 15 min. To obtain protein-free extracts, samples were transferred to Amicon Ultra concentrators (molecular mass cutoff of 10 kDa, Millipore) and centrifuged at 10,000 ϫ g for 15 min. The flow-through was collected, and AMP was quantified using the published procedure (27).
In Vitro tRNA Synthesis-For quantification of tRNA thiolation by the URM1 variants, the specific tRNA for lysine (UUU) was synthesized in vitro. For synthesis, the following primers were used tRNA Lys(UUU) Fw: 5Ј-TAATACGACTCACTATAG-CCCGGATAGCTCAGTCGGTAGAGCATCAGACTTTTA-ATCTG-3Ј and tRNA Lys(UUU) Rev: 5Ј-CGCCCGAACAGGGA-CTTGAACCCTGGACCCTCAGATTAAAAGTCT-3Ј. Primers were annealed, and single-stranded DNA was synthesized by a reverse transcriptase reaction using the RevertAid TM H minus kit (Fermentas). Single-stranded DNA was purified using the Extract II kit (Macherey & Nagel) which served as a template for in vitro transcription using T7 RNA polymerase. Synthesized tRNA was separated by urea containing PAGE. Gel bands corresponding to tRNA were excised, and tRNA was dissolved in buffer and concentrated by ethanol precipitation.
In Vitro tRNA Thiolation and Nucleoside Analysis-To quantify the ability of URM1 and the variants to transfer sulfur to the wobble uridine of tRNA Lys(UUU) , we incubated 200 g of tRNA Lys(UUU) , 5 M MOCS3, 10 M URM1, 2.5 mM Mg-ATP, and 1 mM sodium sulfide. In addition, 10 M CTU2 was included in the reaction, which was expressed in E. coli with an N-terminal His 6 tag and purified by Ni-NTA chromatography and subsequent size exclusion chromatography. The reaction was carried out in a total volume of 200 l and incubated for 60 min at RT. Afterward, the thiolated tRNA was precipitated with ethanol and resuspended in 200 l of 50 mM sodium acetate (pH 5.3) and 20 mM zinc acetate. 3 units of nuclease P1 was added, and the mixture was incubated overnight at 37°C. Sub-sequently, 10 l of 1 M Tris and 4 units of FastAP (Fermentas) were added, and the mixture was incubated for an additional 16 h at 37°C. Nucleoside analysis was performed as described by Gehrke et al. (30) using a LiChrospher 100 RP 18 column (5 m, 250 ϫ 4.6 mm).
Surface Plasmon Resonance (SPR) Measurements-All bindings experiments were conducted on an SPR-based Biacore TM T200 instrument on CM5 sensor chips at a temperature of 25°C and a flow of 30 l/min using the control T200 software and evaluation T200 software (GE Healthcare). The autosampler rack containing the samples was cooled throughout the entire measurements to 8°C. Immobilization of proteins yield the following response units (RU) per flow cell: BSA, 698 RU; MOCS3 787 RU. As running buffer, 20 mM phosphate, 150 mM NaCl, 0.005% (v/v) Tween 20 (pH 7.4) was used. URM1 and MOCS2A variants with concentrations of 0.16, 0.31, 0.63, 1.25, 2.5, 5, 10, and 20 M were injected for 4.5 min at a flow rate of 30 l/min followed by a 15-min dissociation using the kinject command and regeneration of the sensor surface with 50 mM HCl for 1 min. As a control, BSA was used as ligand. Binding curves were corrected by subtraction of buffer injection curves for both flow cells.
Cellular Localization Studies and FRET Analysis-HEK293 or HeLa cells were grown on poly-L-lysine-coated coverslips. Transient transfection of the cells was performed using a modified calcium phosphate method. The DNA/calcium mixture was added dropwise to the cells, and the medium was replaced with fresh culture medium after 8 h. 16 -20 h after transfection, cells were fixed for 30 min using 4% paraformaldehyde in PBS at 4°C. The cells were washed twice with PBS and mounted onto slides with Mowiol (Roth). Images for subcellular localization were conducted at a Zeiss LSM710 (Zeiss, Jena, Germany) laser scanning confocal microscope equipped with a PlanApo 1.4/ 63ϫ objective.
For fluorescence lifetime imaging microscopy, a Visitron Systems imaging system based on an inverted microscope (Axio Observer, Zeiss) equipped with an Infinity 3 confocal scanning head and an additional total internal reflection port were used. As excitation light source, a mode locked ps Nd-YAG laser with regenerative amplifier (PL2201/TH, Ekspla) was used. The third harmonic output at 355 nm and a repetition rate of 1 kHz was 30 J/pulse. The output of the Nd-YAG laser was coupled free space without collimation optics into a 500-m quartz fiber bundle. This fiber bundle was connected to the total internal reflection port at the rear side of the microscope via a 355-nm dichroic mirror. Inside the microscope, an arrangement consisting of an additional 355-nm dichroic filter and a 400-nm cutoff filter was used for the separation of excitation and fluorescence emission light. To spectrally separate the ECFP and EYFP emissions, a dual view module (MAG Biosystems) was placed between microscope and detector. This module was equipped with a dichroic filter set for 485 and 540 nm, respectively, with a 30-nm bandpass for each channel.
For time-gated detection, an iCCD camera (Pimax2, Princeton Instruments) was used. For all measurements, 100 frames with a time increment of 0.25 and 1.8 ns gate width were acquired to monitor an overall time interval of 25 ns. For each single frame, 250 pulses were integrated on a chip. To synchro-nize the iCCD camera to the laser, the Q-switch monitor output of the laser was used to trigger the TTL input of the CCD control unit. Confocal images were acquired with a CoolSnap HQ 2 CCD camera (Photometrics) attached to the Infinity 3 confocal scanning head. Samples were excited with an Ar-Kr ion laser (Innova C70, Coherent) at 456 nm for ECFP or 514 nm for EYFP. Data acquisition and processing were performed with the Metamorph 7.2 software (Molecular Devices). At least 40 cells for each sample were measured. The efficiency of FRET was determined by measuring the donor lifetime of ECFP alone ( D ) and in the presence of an acceptor ( DA ).

Expression and Purification of Human MOCS3 from Sf9 Cells-
Earlier attempts to purify the holo-MOCS3 protein in an active form failed after expression in heterologous systems like E. coli, Pichia pastoris, or S. cerevisiae. Here, an approach was chosen to express human MOCS3 in a baculovirus insect cell system in Sf9 cells. The MOCS3 gene was cloned into the pFastBac1 vector resulting in a fusion of the expressed recombinant protein with an N-terminal His 6 tag. The Sf9 cells were grown as suspension cultures transfected at a density of 1 ϫ 10 6 cells/ml with recombinant virus and harvested by centrifugation 72 h post-infection. The soluble fraction of the cell lysate containing MOCS3 was purified by Ni-NTA chromatography. After elution, one major band was detected on Coomassie Brilliant Blue R stained SDS-polyacrylamide gels with a size of ϳ52 kDa (Fig.  1A), corresponding to the calculated molecular mass of 56,695 Da for His 6 -tagged MOCS3. The protein was purified with a yield of 1.4 mg/100 ml of Sf9 cells and a purity of more than 90%.
Analysis of the Thiosulfate:Sulfur Transfer Activity of MOCS3-Previous studies using the single C-terminal rhodanese-like domain of MOCS3 (MOCS3-RLD) demonstrated that in an in vitro reaction the protein transfers sulfur from thiosulfate to cyanide (10). The sulfur is bound as a persulfide on Cys-412 of the MOCS3-RLD, the first residue of the conserved six-amino acid active site loop characteristic for rhodaneses (10,31). Here, we compared the thiosulfate sulfurtransferase activity of holo-MOCS3 purified from Sf9 cells to the activity of its separately purified RLD expressed in E. coli cells. The kinetic constants are listed in Table 1 and show that holo-MOCS3 and MOCS3-RLD have comparable k cat and K m values, showing that MOCS3 was purified with a functional C-terminal RLD from Sf9 cells. The low catalytic efficiency with thiosulfate as sulfur source shows that thiosulfate is likely not the physiological substrate of MOCS3, as already described by Krepinsky and Leimkühler (31) in a study using the MOCS3-RLD protein. In vivo, the sulfur donor for MOCS3 was proposed to be the L-cysteine desulfurase NFS1 (11), a protein also involved in FeS cluster biogenesis.
Site-directed Mutagenesis of MOCS2A and URM1-The amino acid sequence comparison of URM1 and MOCS2A showed an identity of 24%, with the most striking feature being the C-terminal double glycine motif (Fig. 2). To compare the role of the C-terminal Gly-Gly motif of MOCS2A and URM1 for the interaction with MOCS3, the following single amino acid variants were generated and purified: URM1: G101A, G101⌬, and 102Gϩ; MOCS2A: G88A, G88⌬, and 89Gϩ.
The proteins were expressed in a heterologous expression system in E. coli and purified by ammonium sulfate precipitation and size exclusion chromatography (URM1) or by chitin affinity chromatography (MOCS2A). All variants exhibited expression and purification characteristics similar to those of the wild type proteins, showing that the amino acid changes did not cause significant structural changes that result in protein instability or altered solubility (Fig. 1, B and C). The proteins were purified with a yield of ϳ3 mg/liter of expression culture and a purity of at least 95%. For both URM1 and MOCS2A, two bands were obtained after purification (Fig. 1, B and C). By MS/MS peptide mapping, we determined that both bands corresponded to URM1 or MOCS2A, respectively, most likely reflecting the presence of the adenylated and nonadenylated species as shown before for E. coli MoaD (32).
Analysis of the MOCS3-catalyzed Adenylation of MOCS2A and URM1 Variants-The first step of activation of both MOCS2A and URM1 consists of the formation of an acyl-adenylate group at the C-terminal glycine residue. To test the activity of the N-terminal MoeB-like domain of purified MOCS3, the adenylation efficiency of MOCS3 was analyzed with MOCS2A and URM1 wild type in addition to the generated protein variants. The incubation mixtures contained MOCS3, Mg-ATP, inorganic pyrophosphatase, and either MOCS2A or URM1 variants. The reaction was stopped after  varying time points by the addition of 1% SDS and a heat inactivation step. The released AMP was separated on an HPLC reversed phase column and detected by its absorption at 260 nm. A maximum rate of AMP production (1 mol of AMP/mol of protein) was obtained after 90 min for the wild type proteins.
The results in Fig. 3 show that the C-terminal glycine is essential for the adenylation of both MOCS2A and URM1, because a deletion of the last glycine resulted in a complete loss of AMP formation. The variation of the last glycine of both of URM1 and MOCS2A resulted in reduced AMP levels. However, the effect in MOCS2A-G88A was more pronounced with 67% reduction in AMP production in comparison with Urm1-G101A, where the level was only 23% reduced. When an additional glycine was added, the adenylation efficiency in URM1 was reduced to 40% and in MOCS2A to 23%. Overall, the amino acid variations had the same effect on the adenylation efficiency of both proteins, and in MOCS2A, only the amino acid exchanges had a higher impact on the overall reaction. Analysis of the Rate of MPT Formation by MOCS2A-After formation of the acyl-adenylate on the C terminus of MOCS2A, sulfur is transferred from the C-terminal MOCS3-RLD, and a thiocarboxylate group is formed. In case of MOCS2A, two thiocarboxylated proteins then transfer the sulfur to MOCS2Bbound cPMP, and MPT is released. To determine whether the alterations at the C terminus of MOCS2A affected MPT formation, a defined in vitro system was used consisting of purified cPMP, MOCS3, E. coli MoaE, Mg-ATP, inorganic pyrophosphatase, and MOCS2A. We substituted MOCS2B by E. coli MoaE in our assay, which was shown to be more active in vitro (8). As sulfur source, either sodium sulfide or thiosulfate were used. Sulfide can be added to activated MOCS2A-AMP without the activity of the C-terminal RLD of MOCS3, whereas thiosulfate is converted to sulfite and a protein-bound persulfide on MOCS3-RLD, which is then transferred further on to MOCS2A. Like this, the activity of the RLD can be determined and compared with the direct sulfur transfer with sulfide in vitro. For quantification, produced MPT was converted to its stable oxidation product Form A by oxidation with acidic iodine. The reaction mixtures were separated on a reversedphase HPLC column, and Form A was quantified by its fluorescence. As shown in Fig. 4 (white bars), when sulfide was used as sulfur source, the effect on MPT production of the MOCS2A variants correlated to the adenylation efficiency shown above. The MOCS2A-G88⌬ variant showed a level of 5% MPT pro-duced in the assay in comparison with wild type MOCS2A. When using the MOCS2A-89Gϩ and MOCS2A-G88A variants, the MPT levels were reduced to 39 or 59%, respectively. When thiosulfate was used as a sulfur source, the results obtained were comparable (Fig. 4, black bars); however, the overall reaction was less effective and showed a reduction of   MPT production of 25% under the assay conditions. This might imply that only 75% of the purified MOCS3 has an active RLD. In total, the results are highly consistent with the adenylation level of the MOCS2A variants and show that thiocarboxylate formation and adenylation of MOCS2A correlate.
Quantification of tRNA Thiolation Efficiency by URM1-URM1 transfers the sulfur of the thiocarboxylate group to the wobble uridine 34 of tRNAs, namely tRNA Lys (UUU) , tRNA Gln (UUG) , and tRNA Glu (UUC) . The introduction of sulfur at the 2-position results in the formation of s 2 U or mcm 5 s 2 U. Prior to the sulfuration reaction, the uridine is proposed to be activated in an ATP-dependent manner by the CTU1-CTU2 complex (33). It was shown previously (16) that in a defined in vitro system using isolated tRNA Lys(UUU) , URM1, Uba4, and Ncs6 and the yeast homologues of MOCS3 and CTU1, 2-thiouridine was readily formed. The tRNA was digested; nucleosides were separated, and formed thionucleosides were quantified. Using this procedure, we analyzed the effects of the alterations of the Gly-Gly motif of URM1 on the rate of 2-thiouridine formation. The results shown in Fig. 5 reveal that a drastic reduction of s 2 U was observed for the URM1-G101⌬ variant, whereas for the URM1-G101A and URM1-102Gϩ variants, s 2 U was detected at a level of 58 and 48%, respectively, in comparison with wild type URM1 (Fig. 5). The results are highly consistent with the adenylation level of the URM1 variants and show that thiocarboxylate formation and adenylation of URM1 correlate.
Analysis of Protein-Protein Interactions by Surface Plasmon Resonance Measurements-To analyze the influence on the C-terminal glycine of MOCS2A and URM1 on the interaction and complex formation with MOCS3 in vitro, SPR measurements were employed for real time detection of the specific interactions using the purified proteins.
Only in the presence of 100 M ATP in the running buffer was an interaction of MOCS3 with the wild type MOCS2A or URM1 proteins determined. K D values of 0.26 M for URM1-MOCS3 and 0.47 M for MOCS2A-MOCS3 were determined using a 1:1 binding model. In case of URM1-G101A, binding curves were monitored; however, because the response did not correspond to a 1:1 binding model, a K D value could not be calculated. For the other variants, no apparent interactions with the immobilized MOCS3 were detectable. The results show that the effects of the variation of the C-terminal glycine of MOCS2A and URM1 variants observed above are based on an impaired ability of the proteins to form a stable complex with MOCS3. SPR was unable to resolve the differences in the interaction between MOCS3 and alteration of the C terminus of MOCS2A and URM1, which were detected in the activity assays. Additionally, the results show that the interaction of MOCS2A or URM1 with MOCS3 is more stable in the presence of ATP, as reported before for the Ub-E1 interaction (34).

Analysis of the Effects of MOCS2A and URM1 Variants on Their Cellular Localization and the Interaction with MOCS3 in
Human Cells-To further investigate the role of the C-terminal glycine of MOCS2A and URM1 in vivo, the variants were expressed as an N-terminal EYFP fusion in HeLa cells. As reported previously (10), our investigations determined that MOCS3 was located in the cytoplasm, whereas MOCS2A showed a dual localization in the cytoplasm and the nucleus (Fig. 6A). URM1 showed a similar distribution like MOCS2A, with a localization in the nucleus and in the cytoplasm of HeLa and HEK293 cells (Fig. 6B, data not shown). In the localization studies, we identified a correlation between the length of the Gly-Gly motif and the subcellular localization of the two proteins. When the C-terminal glycine of URM1 was deleted, the overexpressed protein was solely identified in the cytosol (Fig.  6B). In addition, the MOCS2A-G88⌬ variant showed lower levels of fluorescence within the nucleus compared with the cytoplasm, which likely corresponds to lower levels in protein amount (Fig. 6A). A similar pattern was observed for the expression of MOCS2A-G88A (Fig. 6A). In case of the extension of the Gly-Gly motif by one residue, both Ubls exhibited expression like the wild type proteins, with a localization both in the cytosol and the nucleus. Surprisingly, when MOCS3 was co-expressed with MOCS2A-89Gϩ or URM1-101Gϩ, MOCS3 was additionally located in the nucleus (Fig. 6B), revealing an influence not only on the localization of the Ubls themselves but also in addition to the activating enzyme.
Furthermore, we also determined the cellular in vivo interaction of MOCS3 with either MOCS2A or URM1 by analyzing their fluorescence resonance energy transfer (FRET). Either N-terminally tagged ECFP or EYFP variants were used, and the FRET efficiency was calculated by determining the decrease in the ECFP donor lifetime. The lifetime was determined by fluorescence lifetime imaging microscopy measurements with an excitation of ECFP at 355 nm and the decay curves were measured in an overall time interval of 25 ns. The decay curves were fitted bi-exponentially with a long lifetime component of 1 ϭ D ϭ 3.9 Ϯ 0.1 ns and a short lifetime of 2 ϭ 1.2 Ϯ 0.1 ns in the shorter range, which corresponds to an internal conformational change in ECFP as reported before (35). As positive control, a protein fusion of ECFP and EYFP with peptide linker was used, for which the donor long lifetime component was reduced to 1 ϭ DA ϭ 2.8 Ϯ 0.1 ns, although the short lifetime of 2 ϭ 1.2 Ϯ 0.1 ns remained constant (Fig. 7). This corresponds to a decrease in the lifetime of 26%. In comparison, the co-expression of ECFP and EYFP alone without the peptide linker did not alter the donor lifetime (Fig. 7). The FRET resulting from the co-expression of MOCS3 with wild type MOCS2A and URM1 resulted in a reduction of 20% in the donor lifetime, due to a higher distance between the fluorophores in the heteromeric complex compared with the distance in the ECFP-EYFP fusion protein. When EYFP was fused to the C terminus of either URM1 or MOCS2A, the ECFP-MOCS3 fusion did not change its lifetime, showing that the proteins were unable to interact due to the C-terminal EYFP location.
The resulting lifetimes of the N-terminal EYFP-MOCS2A and URM1 fusions with ECFP-MOCS3 are shown in Fig. 7. The results show that replacement of the ultimate glycine of URM1 by alanine resulted in a minor decrease in donor lifetime, whereas deletion or addition of a glycine to URM1 resulted in a drastic decrease in the donor lifetime (Fig. 7B). Thus, the proteins were not able to efficiently interact in vivo.
In case of MOCS2A, the alteration of the ultimate glycine in the MOCS2A-G88A variant resulted in reduced FRET efficiencies, whereas the deletion of the ultimate residue in MOCS2A-G88⌬ abolished the energy transfer from ECFP completely (Fig.  7C). In contrast, in MOCS2A-89Gϩ, the addition of a C-terminal glycine resulted in a decrease in lifetime that was not as pronounced as for the URM1-102Gϩ variant. In total, the results are highly consistent with the obtained activities for the protein variants and show that the interaction of both proteins is primarily determined by the C-terminal glycine of the ␤-grasp fold protein.

DISCUSSION
In this report, we expressed and purified MOCS3 in a baculovirus insect cell system from Sf9 cells, which so far is the only suitable system for the expression of MOCS3. All other attempts to express MOCS3 in eukaryotic systems like P. pastoris or S. cerevisiae did not yield an active enzyme. Thus, for the first time we were able to demonstrate that MOCS3 was purified in an active form from Sf9 cells, and we demonstrated that both the N-terminal E1/MoeB-like domain and the C-terminal RLD are active. The first indications that MOCS3 might interact with both URM1 and MOCS2A were obtained from in vitro studies in which MOCS3 was replaced by the S. cerevisiae homologue Uba4 (5). The in vitro interaction studies showed that Uba4 was able to interact with both human MOCS2A and human URM1. In addition to the interaction, in these studies we were able to demonstrate that a thiocarboxylate group is formed on URM1 in vitro. Later, the thiocarboxylate on Urm1 was confirmed by in vivo studies (23). Even though S. cerevisiae harbors both URM1-mediated tRNA thiolation and its protein conjugation to Ahp1p, it is not an ideal model system, because the genes for Moco biosynthesis, including MOCS2A are absent in this organism. Thus, S. cerevisiae is not a suitable model organism to prove a shared pathway for Moco biosynthesis and tRNA thiolation.
To investigate the role of MOCS3 as a partner for both MOCS2A and URM1 in human cells, we performed FRET studies and investigated the donor lifetime as an indication for the strength of the interaction of the protein pairs. MOCS3 and either MOCS2A or URM1 were co-expressed as ECFP/EYFP fusion proteins in HeLa cells to determine the FRET between the fluorescent dyes. Because of the high dependence of the distance between the dyes, only the complex formation between MOCS3 and the candidates enable an energy transfer. Our results obtained from the donor lifetime measurements showed that MOCS3 is able to interact with both MOCS2A and URM1 in HeLa cells. The same results were additionally obtained in HEK 293 cells (data not shown). Because a similar decay in donor lifetime was obtained, this also revealed a similar distance between the proteins, reflecting similar interaction sites.
For the activation reaction, MOCS3 has to form a heterotetrameric complex with URM1 or MOCS2A, as described previously for Uba4 (5) or the E. coli congeners MoaD and MoeB (4). The crystal structure of the E. coli (MoaD-MoeB) 2 complex showed that it is composed of a central MoeB dimer with a MoaD subunit at each end of the dimer. The C terminus (residues 76 -81) is extended into the pocket on the MoeB surface (4) and is therefore accessible for activation. The yeast Urm1 crystal structure is very similar to the MoaD structure that has two additional ␣-helices, as compared with other Ubls (20), and thus it is more similar to MoaD in this respect. The structure of yeast Urm1 revealed a hydrophobic patch on the surface that is similar to MoaD, and because this region is exposed it might be essential for interaction with MOCS3 (20). Overall, URM1 and MOCS2A share an amino acid sequence identity of 24% (Fig. 2). The double glycine motif at their C termini is the only striking feature in both proteins. To investigate the role of the C-terminal glycine for both proteins for the interaction with MOCS3 in addition to their reactivity, several amino acid variations at the C-terminal glycine were generated. The main purpose was to identify differences between both proteins, because URM1 is a dual function protein and acts in protein conjugation and as a sulfur carrier, and the role of MOCS2A is restricted to act as a sulfur carrier in Moco biosynthesis.
The C-terminal glycine is involved in the first activation step of MOCS2A or URM1 that is catalyzed by MOCS3, involving the formation of an acyl-adenylate as an essential prerequisite for subsequent formation of the thiocarboxylate (5). Our results show that the alteration of the terminal glycine motif had the same effects on URM1 and MOCS2A, because all variants showed an impaired rate of adenylation. In addition, also the variation of the length of the glycine motif had similar effects in both Ubls. The deletion of the C-terminal residue totally abolished adenylation, whereas the extension of the motif by one glycine solely reduced the efficiency of the acyl-adenylate formation. The glycine to alanine exchange of the ultimate glycine of URM1 resulted in a slightly impaired ability to form an acyladenylate; in contrast, the adenylation efficiency for MOCS2A was diminished to about a third compared with the wild type protein.
Considering that the formation of the acyl-adenylate is only the first step of the ATP-dependent activation catalyzed by MOCS3, which is followed by the subsequent thiocarboxylation catalyzed by the RLD, we investigated the sulfur transfer reaction of thiocarboxylated MOCS2A to cPMP. In our studies, MOCS3 was able to thiocarboxylate MOCS2A and the variants using either sulfide or thiosulfate, showing that both domains of MOCS3 are at least 75% active (31). To determine the sulfur transfer rate, MPT formation was monitored. The results showed comparable values as obtained for the rate of adenylation of the proteins, with the MOCS2A-G88⌬ variant having the most prominent effect (23). Similarly, we investigated the transfer of sulfur to the wobble uridine of the tRNA Lys(UUU) by URM1 and the variants. In an in vitro system consisting of MOCS3, CTU2, Mg-ATP, and sodium sulfide, URM1 was able to catalyze 2-thiouridine formation. The exchange of the C-terminal glycine to alanine and the addition of a glycine decreased the rate of 2-thiouridine formation to 50%, although the deletion of the last glycine had a more severe impact showing only 10% of thiouridine. These results are in good agreement to the findings of van der Veen et al. (23) where no thiocarboxylation of URM1-G101⌬ could be detected in vivo, leading to an increase in hypomodified tRNA. Similar findings were also obtained by Noma et al. (15) using the yeast Urm1-⌬GG variant, illustrating that deletion of both glycines had similar effects in comparison with the deletion of only the last glycine.
All effects of the variation of the Gly-Gly motif are based on insufficient complex formation with MOCS3, as shown in vitro by SPR measurements. For the wild type proteins in presence of ATP, K D values in the nanomolar range were obtained. In comparison for the variants, only the URM1-G101A showed an impaired ability to form a complex, which did not obey a 1:1 binding and therefore did not allow the determination of a K D value. The FRET studies revealed comparable effects in vivo that could be quantified, showing only a slightly impaired donor lifetime decrease for the variants compared with the more profound effects in case of the wild type proteins.
Our results are also in good agreement with a previous study investigating the role of the glycine motif of E. coli MoaD (27). These studies together underline the assumption that the ultimate glycine is optimized for the sulfur carrier role for Moco biosynthesis and tRNA thiolation and that especially the length of the tail of the Ubl is critical, because deletion of the tail generally resulted in a inactive protein for all Ubls (27). In contrast, the ubiquitin-G76A variant exhibited quantitatively acyl-adenylate and thioester formation with its E1 (36), which is consistent with our findings for the URM1-G101A variant, which also showed significant activities in acyl-adenylate formation. This implies that URM1 is more similar to the protein conjugation systems than MOCS2A, which is optimized to serve only as a sulfur carrier protein.
Surprisingly, while determining the FRET efficiency of URM1 and MOCS2A with MOCS3, an influence on the localization by the different variations was observed, revealing a connection between the alteration of the Gly-Gly motif and the subcellular localization. Generally, MOCS3 has a cytosolic localization, although URM1 and MOCS2A are both found in the cytosol and the nucleus. The different modifications at the C terminus of the Ubl affected the protein localization to the same extent, revealing a restricted localization to the cytosol for MOCS2A-G88A, MOCS2A-G88⌬, and URM1-G101⌬. In addition, the co-expression of MOCS3 with the variants of URM1 and MOCS2A with the C-terminal elongation of the Gly-Gly motif by one glycine showed not only a nuclear localization, like observed for the wild type proteins, but also an influence on the localization of their E1 partner, MOCS3, which exhibited an additional localization inside the nucleus. So far, it is unclear whether URM1 or MOCS2A has a specific role in the nucleus.
Very recently van der Veen et al. (23) identified targets of urmylation in humans, among which were parts of the tRNA thiolation machinery (CTU1 and CTU2) and a deubiquitinating enzyme (USP15), proteins involved in nuclear transport like CAS, and a protein involved in the shuttling between cytosol and nucleus. Taking into account that upon urmylation CAS was relocalized to the cytosol and the previous description of the influence of Ubl conjugation on the subcellular localization of proteins (23), there might be an effect on the localization of MOCS3 by its substrates.
Van der Veen et al. (23) additionally described the role of the nucleocytoplasmic shuttling factor CAS as a target of urmyla-tion linking the nuclear transport to oxidative stress, which is in good agreement with the observation that nuclear import is haltered upon oxidant treatment as also reported by Kodiha et al. (37). Speculation on the influence of urmylation on the arrest of nuclear transport is underlined by similar phenotypes of CAS depletion (38) and URM1 silencing (13), which both result in the arrest in G 2 /M phase.
However, the morphological abnormalities caused by reduction of cellular URM1 levels in HeLa cells by shRNA, which were shown to be increased cell size and the presence of multiple nuclei within a single cell, likely also result from translational errors, because the 2-thio modification is implicated in translational fidelity and efficacy as stated by Ikeuchi et al. (39). Additionally, similar findings were also reported for MOCS3 depletion in Caenorhabditis elegans (40). Here, depletion of the MOCS3 homologue reduced the levels of tRNA thiolation leading to misreading of an opal stop codon and subsequently the suppression of a multivulva phenotype, demonstrating the role of tRNA modification on global cellular processes in this organism. Furthermore, multinuclear cells were also observed after overexpression of URM1 in human cell cultures. These findings demonstrate an additional role of URM1 for cytokinesis and cell cycle progression. Taken together with the results presented herein, a contribution of URM1/MOCS3 to the control of global processes like nuclear transport, cytokinesis, and cell cycle progression is possible. In comparison, yeast MDY2, a Ubl domain protein with no exposed C-terminal Gly-Gly motif, is associated to microtubule-directed nuclear migration, which might have additionally a potential role in these pathways, because MOCS3 was localized in centrosomes where the microtubule emanated. For MDY2, it was described that it is mostly localized in the nucleus, but upon heat stress the protein is relocalized to cytosolic granules. In MYD2, there is no evidence for C-terminal processing, but it is involved in nuclear migration via the microtubules, which emanates from spindle pole body, the yeast equivalent of the centrosome (41). For SUMO, it has been described that it is required for the recruitment of RanGAP1 to kinetochores and the mitotic spindle to control processes during mitosis (42), and it has an essential role in the control of nuclear protein import (43). Thus, MOCS3 might be shuttled into the nucleus by either URM1 or MOCS2A. However, whether MOCS3 has a role in the nucleus still has to be investigated.
Our data clearly show that MOCS3 marks the intersection of two pathways, tRNA thiolation and Moco biosynthesis, by activating both URM1 and MOCS2A in human cells (Fig. 8). Thus, MOCS3 might be the central switch to determine the faith of the persulfide sulfur at its C-terminal RLD. Because MOCS3 was also shown to be a target for urmylation (23), this might have a regulatory role on the activity of MOCS3 and its interaction with either URM1 or MOCS2A. The influence of urmylation on enzyme activity has to be investigated in future studies.