Post-translational formylglycine modification of bacterial sulfatases by the radical S-adenosylmethionine protein AtsB.

C(alpha)-Formylglycine (FGly) is the catalytic residue of sulfatases. FGly is generated by post-translational modification of a cysteine (prokaryotes and eukaryotes) or serine (prokaryotes) located in a conserved (C/S)XPXR motif. AtsB of Klebsiella pneumoniae is directly involved in FGly generation from serine. AtsB is predicted to belong to the newly discovered radical S-adenosylmethionine (SAM) superfamily. By in vivo and in vitro studies we show that SAM is the critical co-factor for formation of a functional AtsB.SAM.sulfatase complex and for FGly formation by AtsB. The SAM-binding site of AtsB involves (83)GGE(85) and possibly also a juxtaposed FeS center coordinated by Cys(39) and Cys(42), as indicated by alanine scanning mutagenesis. Mutation of these and other conserved cysteines as well as treatment with metal chelators fully impaired FGly formation, indicating that all three predicted FeS centers are crucial for AtsB function. It is concluded that AtsB oxidizes serine to FGly by a radical mechanism that is initiated through reductive cleavage of SAM, thereby generating the highly oxidizing deoxyadenosyl radical, which abstracts a hydrogen from the serine-C(beta)H(2)-OH side chain.

C ␣ -Formylglycine (FGly) is the catalytic residue of sulfatases. FGly is generated by post-translational modification of a cysteine (prokaryotes and eukaryotes) or serine (prokaryotes) located in a conserved (C/S)XPXR motif. AtsB of Klebsiella pneumoniae is directly involved in FGly generation from serine. AtsB is predicted to belong to the newly discovered radical S-adenosylmethionine (SAM) superfamily. By in vivo and in vitro studies we show that SAM is the critical co-factor for formation of a functional AtsB⅐SAM⅐sulfatase complex and for FGly formation by AtsB. The SAM-binding site of AtsB involves 83 GGE 85 and possibly also a juxtaposed FeS center coordinated by Cys 39 and Cys 42 , as indicated by alanine scanning mutagenesis. Mutation of these and other conserved cysteines as well as treatment with metal chelators fully impaired FGly formation, indicating that all three predicted FeS centers are crucial for AtsB function. It is concluded that AtsB oxidizes serine to FGly by a radical mechanism that is initiated through reductive cleavage of SAM, thereby generating the highly oxidizing deoxyadenosyl radical, which abstracts a hydrogen from the serine-C ␤ H 2 -OH side chain.
C ␣ -Formylglycine (FGly) 1 is a unique amino acid so far found exclusively in sulfatases where it forms the catalytic residue at the active site (1)(2)(3)(4). In the ground state of the sulfatase, the FGly is hydrated bearing two geminal hydroxyls at the C ␤carbon (5). Upon binding of a sulfate ester substrate, one of the hydroxyls performs a nucleophilic attack on the sulfur atom leading to transfer of the sulfate group onto the enzyme. The second hydroxyl is essential for sulfate elimination from this sulfated enzyme intermediate to regenerate the formyl group (5)(6)(7). In essence, the water molecule initially added to the enzyme for hydration of the formyl group allows performance of "intramolecular hydrolysis" of a sulfate ester. This novel hydrolytic mechanism explains the functional importance of the FGly residue. Failure to generate FGly in human leads to multiple sulfatase deficiency, a rare but fatal inherited disorder characterized by the synthesis of catalytically inactive sulfatases (8 -10).
In most eukaryotes FGly residues are generated by the recently discovered FGly-generating enzyme (FGE), and it is the genetic deficiency of FGE that causes multiple sulfatase deficiency (9,10). FGE acts in the lumen of the endoplasmic reticulum shortly after import of the nascent sulfatase polypeptide and modifies a cysteine residue located in a conserved CXPXR motif (11)(12)(13)(14). This cysteine undergoes C␤-oxidation and Selimination by an as yet unknown reaction mechanism. FGE is encoded in the SUMF1 gene and belongs to a protein family consisting of a domain of unknown function that does not allow prediction of any aspect of its mechanism of action (15). It was proposed that a cysteine cluster, located at the highly conserved C terminus of FGE, may be involved in catalysis, because this region is a hot spot for mutations in multiple sulfatase deficiency patients (9).
SUMF1 genes are lacking in Caenorhabditis elegans, Schizosaccharomyces pombe, and Neurospora crassa that are known to express active sulfatases, which indicates the existence of a second FGly-generating system in eukaryotes. In a variety of prokaryotes SUMF1 orthologs can be found in sulfatase operons (15). As in eukaryotes, some bacteria known to generate FGly residues by cysteine modification, e.g. Escherichia coli and Pseudomonas aeruginosa (16), lack SUMF1 orthologs. Thus, in bacteria a second cysteine-modifying system also exists, orthologs of which may be operative in those eukaryotes lacking FGE, thereby functionally closing the phylogenetic gap mentioned above. It should be noted that FGE as well as the E. coli and P. aeruginosa cysteine-modifying systems are strictly cysteine-specific (6,16).
The remarkable versatility of bacteria to generate the critical FGly residue is highlighted by the fact that a number of species express sulfatases in which a serine residue is oxidized to FGly, a modification directed by a homologous SXPXR motif (17). These serine-type sulfatases, in contrast to the cytosolic cysteine-type sulfatases, show periplasmic localization (18,19) and are specifically modified by a cytosolic protein termed AtsB in Klebsiella pneumoniae (19,20). AtsB is co-expressed with the sulfatase AtsA (arylsulfatase of K. pneumoniae) being encoded on the same ats operon. Co-expression of AtsB was shown to be strictly required for AtsA modification in E. coli where it acts in trans on AtsA (20). AtsB interacts with the FGly modification motif of AtsA in a serine-specific manner, as was shown in yeast two-hybrid and GST pull-down experiments (19). FGly generation from cysteine and serine share the (C/S)XPXR modification motif. Furthermore, in vivo FGE and AtsB both rely on transmembrane translocation of the sulfatase or at least on the presence of a signal peptide in the sulfatase substrate (11,19). Thus, AtsB-mediated FGly generation may occur, when the unfolded sulfatase protein is being translocated, acting on the cis-side of the membrane, in contrast to the microsomal FGE, which acts on the trans-side.
When studying AtsA modification in E. coli, its two chromo-somal atsB orthologs did not substitute for Klebsiella atsB, most likely because bacterial sulfatase operons generally are repressed under conditions of sulfur supply. On the contrary, the cysteine-modifying system of E. coli was found to be expressed under any conditions tested. This difference in the regulation of the structural and modifying genes of the sulfatases may suggest that at least the cysteine-modifying system of E. coli fulfills also further, possibly essential functions and that FGly is present also in some non-sulfatase proteins.
In contrast to FGly generation from cysteine, where no mechanistic information for the modification reaction is available, some aspects of AtsB-mediated serine oxidation may be inferred by bioinformatics. AtsB was proposed to belong to the so-called radical SAM protein superfamily that generates a deoxyadenosyl radical via reductive cleavage of S-adenosylmethionine (SAM) (21). This radical in turn could generate a substrate radical that undergoes further reaction (22)(23)(24). Accordingly, the oxidation of serine to FGly, i.e. serine semialdehyde, would involve two single electron transfer steps. These electrons are likely to be donated/accepted by FeS centers, three of which have been predicted for AtsB on the basis of its conserved cysteine clusters (25). Here we experimentally test this hypothetical mechanism, investigating the involvement of SAM as a co-factor and of the predicted FeS centers as redox catalysts for FGly generation in vivo and in vitro. For construction of GST-AtsA, AtsA-residues 21-112 were fused with GST using the pGEX-KG vector, as described earlier (19). All of the mutations and constructs were analyzed by full-length sequencing of coding regions to preclude any PCR-derived errors.

Materials
Protein Expression-E. coli DH5␣ was transformed with the following plasmids (for construction see above): pGEX-KG containing either gst-atsA or gst-atsA-S72A/P74A/R76A or pBluescript II containing either atsA, atsB, or atsBA (atsB with or without mutations, as indicated for each experiment). SAM-binding site mutants of AtsB were co-expressed with GroEL/ES in double transformants containing also the pT-groE plasmid (26). The presence of the two plasmids was maintained in selective medium containing ampicillin and chloramphenicol and was routinely checked by PCR analysis. Overnight cultures were used to inoculate 100 ml of fresh liquid medium (1% inoculum). For induction of protein expression, isopropyl-␤-D-thiogalactopyranoside was added to 0.2 (for pGEX) or 0.5 mM (for pBluescript II) upon reaching an A 600 value of 0.3-0.5, and culture was maintained for another 3 h at 37°C. For SAM-binding site mutants of AtsB only, expression was performed in overnight (14 h) cultures and inoculated from glycerol stocks without the addition of isopropyl-␤-D-thiogalactopyranoside.
In Vitro Interaction and FGly Modification Assay-Bacteria expressing GST-AtsA were sedimented, washed once in PBS (pH 7.4), resuspended in PBS (pH 7.4) containing 8 M urea, and disrupted in a French press cell. The soluble material (75,000 ϫ g supernatant) was subjected to dialysis against PBS to remove the urea. The dialyzed material was cleared by centrifugation (as above). For in vitro interaction, 0.75 ml of the supernatant was mixed with 0.75 ml of an E. coli French press lysate (in PBS), containing AtsB protein, and incubated at 37°C for 15-60 min in the absence or presence of different concentrations of SAM. Where indicated, [ 3 H]SAM (1 Ci) was added as tracer. After incubation, the interaction mixture, corresponding to ϳ5 mg of total E. coli protein, was loaded on a GSH-agarose column (0.2 ml of bed volume), and the flow-through was immediately collected without further incubation. The column was washed six times with 4 column volumes of PBS and then eluted three times with 1 column volume of 10 mM glutathione in PBS (pH 8.0). The eluate fractions were analyzed for GST-AtsA and AtsB by Western blotting and for [ 3 H]SAM by liquid scintillation counting. To test for in vitro FGly generation, the incubation of the interaction mixture described above was carried out at 37°C for 1-3 h in the presence of 500 M SAM. Where indicated (see Figs. 3, 4C, and 5C), another 500 M SAM was added to the interaction mixture when half of the incubation time had passed. After incubation, the GST-AtsA fusion protein was purified by GSH-agarose, separated by SDS-PAGE, and subjected to tryptic digestion and mass spectrometry (see below). It should be noted that all experiments could also be performed using purified GST-AtsA for interaction and modification reactions. However, using purified AtsB led to poor interaction with GST-AtsA. Rather, the E. coli lysate containing AtsB had to be freshly prepared avoiding exposure to room temperature.
Protein and Peptide Analysis-AtsA, AtsB, or GST fusion proteins were detected by Western blotting using anti-AtsA, anti-AtsB, or anti-GST as primary antibodies. ECL signals of corresponding secondary antibodies were detected by a LAS1000ϩ imaging system (Raytest) and quantitated by densitometry of digital images using Aida 3.10 software (Raytest). For SDS-PAGE see Ref. 27. The sulfatase activity of expressed AtsA was determined in duplicate assays at saturating substrate concentration, as described earlier (17).
The presence of FGly in AtsA and GST-AtsA was analyzed by MALDI-TOF mass spectrometry at the level of their tryptic peptides. The proteins were separated by SDS-PAGE and stained with Coomassie Brilliant Blue. From the gels, the stained AtsA or GST-AtsA bands were excised and subjected to in-gel digestion with trypsin and extraction of tryptic peptides according to standard protocols (28). The presence of FGly in the tryptic AtsA peptide 2 (P2) was verified by mass spectrometry on a matrix-assisted laser desorption ionization-time of flight Reflex III instrument (Bruker Daltonics), using as matrix 1.3 mg/ml recrystallized 2,4-dinitrophenyl hydrazine in 50% aqueous acetonitrile solution containing 0.5% trifluoroacetic acid. The samples were prepared by the drying droplet method, drying 0.5 l each of sample and matrix solution on a stainless steel target. For further details and for MALDI post-source decay MS analysis, see the work of Peng et al. (29).

Formation of a Functional Complex of AtsB with AtsA
Requires S-Adenosylmethionine-AtsB has been predicted to belong to the radical SAM superfamily generating substrate-or enzyme-located radicals with the help of SAM (21). To test for a possible involvement of SAM in AtsB-mediated FGly modification of AtsA, we performed in vitro AtsB-AtsA interaction experiments in the absence or presence of SAM or its demethylated analog SAHC. Using an AtsA fragment (amino acid residues 21-112) fused to the C terminus of glutathione S-transferase (GST-AtsA) and an E. coli cell lysate containing overexpressed AtsB, we observed an association of AtsB with GST-AtsA that both bound to and co-eluted from GSH-agarose, as was described earlier (see Ref. 19). It turned out that addition of 500 M SAM significantly improved AtsB binding, whereas 500 M SAHC completely inhibited AtsB binding (Fig.  1A). The association of GST-AtsA and AtsB and its stimulation by SAM required incubation of all interaction partners at 25-37°C for Ն5 min; at 0°C interaction was rather inefficient (Fig. 1B). This suggests that formation of a stable complex, which can be pulled down by GSH-agarose, involves a ratelimiting step that is temperature-dependent. In the presence of SAM, formation of this complex increased with time over a 60-min period (Fig. 1C). Without the addition of SAM, the complex formed initially but was lost upon incubation for longer than 30 min. This transient complex is attributed to the presence of some endogenous SAM in the E. coli lysate, which initially promotes complex formation and decomposes during the 37°C incubation, thus destabilizing the complex. Control experiments showed that GST-AtsA⅐AtsB complex formation in the presence of 50 M SAM reached a maximum after Ϸ15 min at 37°C. At Ն300 M SAM the complex was stable for more than 60 min ( Fig. 1C and data not shown).
To prove that SAM indeed is part of a ternary GST-AtsA⅐AtsB⅐SAM complex, we added 3 H-labeled SAM as a tracer to the incubation mixture containing 50 M SAM. After loading this mixture to GSH-agarose and extensive washing of the column, the GSH eluate was collected and analyzed for AtsB by Western blotting and for SAM by liquid scintillation counting. It turned out that significant amounts of radioactivity bound to the column and co-eluted with AtsB. No radioactivity was recovered in the eluate of GSH-agarose loaded with an incubation mixture lacking AtsB ( Fig. 2A, compare first and third lanes). In a further control GST-AtsA-S72A/P74A/R76A was used as bait. This mutant bound neither AtsB nor SAM ( Fig.  2A, second lane).
SAM stimulated GST-AtsA/AtsB interaction in a concentration-dependent manner. Under standard conditions (incubation for 15 min at 37°C), half-maximum stimulation was observed at ϳ20 -40 M SAM, and saturation was observed at ϳ150 -300 M SAM (Fig. 2B). Also the amount of [ 3 H]SAM co-eluting from GSH-agarose together with GST-AtsA and AtsB showed a clear dependence on the applied SAM concentration. At Ն300 M SAM nearly equimolar amounts of AtsB and SAM were recovered in the eluate (Fig. 2B); in three experiments SAM and AtsB bound in a ratio of 0.69 Ϯ 0.06. At nonsaturating SAM concentrations, the amount of bound SAM was clearly lower than the amount of bound AtsB (Fig. 2B). This difference may reflect a cooperative stimulation of AtsB binding by SAM to avoid deoxyadenosyl radical formation in the absence of substrate (see "Discussion"). However, we cannot exclude that some SAM is lost from the AtsB⅐GST-AtsA complex because of catalytic turnover (see below) and/or because of the non-steady-state conditions on the affinity column during washing and elution. Previous results have shown that AtsB and GST-AtsA associate with each other in a ratio of 0.54 (19). Taken together these data suggest that binding of SAM to AtsB efficiently promotes formation of a ternary AtsA⅐AtsB⅐SAM complex and that the three interaction partners are likely to be present in this complex in a 1:1:1 ratio.
SAM-dependent in Vitro FGly Formation by AtsB-The data shown above suggest that FGly modification of serine depends on SAM. To directly measure the effect of SAM on the FGlygenerating activity of AtsB, an in vitro system with limiting SAM levels had to be established. In a previous study, we failed to observe FGly formation in an in vitro system, when GST-AtsA was incubated with a soluble extract of E. coli overexpressing AtsB (19). Here we used essentially the same components, this time supplementing the modification reaction with 500 M SAM. After incubation at 37°C, GST-AtsA was isolated from the mixture by GSH-agarose and subjected to digestion with trypsin and MALDI-TOF MS analysis using DNP-hydrazine as a matrix. This method efficiently converts the FGlycontaining tryptic peptide of AtsA (1588 Da) into the corresponding hydrazone (1768 Da) upon laser-assisted desorption from a DNP-hydrazine matrix, and the resulting product can be detected with high sensitivity (29). Formation of the FGly- modified peptide of AtsA, detected as the DNP-hydrazone (1768 Da), was indeed observed when the incubation was performed in the presence of SAM (Fig. 3A) but not in its absence (not shown). The molecular identity of the DNP-hydrazone peptide was confirmed by the indicative 2-oxodihydroimidazole-derivatized fragment ion (524 Da; Fig. 3C) desorbing with high efficiency upon MALDI post-source decay analysis of the 1768-Da parent ion (for generation of this fragment from FGlycontaining peptides, see Ref. 29). The formation of FGly was dependent on AtsB and on the FGly modification motif of AtsA. Substitution of Ser 72 by alanine in AtsA prevented FGly formation by AtsB (data not shown). The in vitro modification reaction, however, was found to be rather inefficient (Fig. 3B). It should be noted that the peak heights of the FGly-DNPhydrazone peptide (1768 Da) and of the unmodified peptide (1590 Da) in Fig. 3A cannot be used to quantitate FGly modification, because during MALDI-TOF MS the former peptide desorbs much better from the used DNP-hydrazine matrix than the latter peptide (29). The efficiency of FGly formation could be increased up to 5-fold by repeated addition of SAM during the incubation (Fig. 3B). This effect is attributed to the limited thermostability of SAM. Nevertheless, from a comparison with AtsA of known FGly content (see Ref. 29) we estimate that at best ϳ10% of the GST-AtsA substrate was modified under conditions where GST-AtsA was in a roughly 10-fold excess over AtsB. Apart from the poor efficiency of the in vitro system, the data obtained show unequivocally that AtsB-mediated FGly formation relies on SAM as an essential co-factor. 83 GGE 85 of AtsB Is Required for SAM Binding, AtsA Interaction, and FGly Formation-AtsB comprises the amino acid sequence GGEPLL at positions 83-88 that distantly resembles the glycine-rich SAM-binding site in the majority of non-DNA methyltransferases (30). The sequence perfectly agrees with the semi-consensus sequence found in an alignment block of the radical SAM superfamily (21). Yet it should be noted that the individual sequence of each family member may vary considerably from the semi-consensus sequence.
To study the importance of the putative SAM-binding site for AtsB-mediated FGly formation in vivo, we performed an alanine scanning mutagenesis for residues 83 GGEPLL 88 (Pro 86 additionally was substituted by glycine to allow for enhanced structural flexibility at this position). After co-expression of these mutants with AtsA in E. coli, the specific sulfatase ac-tivity of AtsA was determined as an indirect indicator of the FGly modification activity of AtsB. It turned out that the specific sulfatase activity of AtsA, i.e. the ratio of activity over Western blot signal, was drastically reduced when Gly 83 , Gly 84 , or Glu 85 of AtsB was replaced by alanine (313-, 31-, and 50-fold lower specific activities, respectively, as compared with AtsA co-expressed with wild type AtsB; Fig. 4A). For the other mutants, at positions 86 -88, wild type sulfatase activities were observed. The expression level of AtsB and its solubility were  Fig. 3, A and B) is given as follows: ϩϩϩ, high abundance (signal/noise ratio Ͼ20); ϩ, low abundance (signal/noise ratio, 2-5); Ϫ, no signal. B, in vitro interaction reactions of GST-AtsA and AtsB, expressed in wild type, G83A, G84A, E85A, or P86G mutant form, were incubated for 15 min at 37°C in the presence of 50 M [ 3 H]SAM (cf. Fig. 2A). The components of these reactions binding to and eluting from GSH-agarose were analyzed by Western blotting and liquid scintillation counting. The positions of the detected proteins, the relative amounts of AtsB, detected by quantitation of the Western blot signals, and the amount of SAM-associated radioactivity recovered in a 33% aliquot of the GSH eluate are given. C, for in vitro FGly modification, GST-AtsA and AtsB expressed in wild type, G84A, or P86G mutant form were incubated in the presence of E. coli protein and 500 M SAM or SAHC, as indicated, for 2 h at 37°C, with the readdition of 500 M SAM or SAHC after 1h (cf. Fig.  3B). The presence (ϩ) or absence (Ϫ) of FGly in GST-AtsA is given, which was detected as DNP-hydrazone derivative by MALDI-TOF MS analysis of its tryptic peptides. not affected by the mutations (Fig. 4A).
To directly determine the AtsB-mediated FGly formation, the co-expressed AtsA protein was subjected to FGly analysis using the same semi-quantitative MALDI-TOF MS method described above, which detects the FGly-modified peptide as a DNP-hydrazone derivative of 1768 Da. The intensity of this 1768-Da signal was strongly reduced in the case of AtsA coexpressed with AtsB mutants G84A and E85A and was not at all detectable in case of co-expression with AtsB-G83A (Fig.  4A). Thus, the drastically reduced AtsA activities, caused by these three AtsB mutants, fully agree with the impaired FGly modification of AtsA.
In vitro AtsA-AtsB interaction experiments in the presence of SAM demonstrated that binding of AtsB mutants G83A, G84A, and E85A to GST-AtsA was totally abolished and that binding of the AtsB-P86G mutant was reduced to 60% (Fig.  4B). In the corresponding eluates from the GSH-agarose columns, measurable amounts of [ 3 H]SAM only were detected for the wild type AtsB control, indicating that none of the AtsB mutants tested was able to bind SAM (Fig. 4B).
The AtsB-G84A and P86G mutants, being either fully (G84A) or only partially (P86G) impaired in binding of AtsA and SAM, were selected to directly test their FGly-generating activity using the in vitro assay described above. It turned out that the latter but not the former mutant was able to catalyze FGly generation in an SAM-dependent and SAHC-sensitive manner (Fig. 4C). Thus, although apparently having a clearly reduced affinity for SAM, the AtsB-P86G mutant obviously is catalytically active.
Taken together these findings strongly suggest that 83 GGEP 86 indeed constitutes the SAM-binding site of AtsB, or a part thereof, and that its occupation by SAM is required for the formation of a ternary AtsB⅐SAM⅐AtsA complex that catalyzes the FGly modification of AtsA. Out of this tetrapeptide sequence, 83 GGE 85 is crucial for AtsB function. At position 86, substitution of proline by glycine was tolerated in vivo, although in vitro it partially impaired the formation of the ternary complex.
Requirement of Conserved Cysteine Residues in AtsB for FGly Generation in Vivo and in Vitro-Radical SAM proteins generate deoxyadenosyl radicals through reductive cleavage of SAM (see the Introduction). Reduction requires an FeS center that is coordinated by a specific cluster of three conserved cysteines (CX 3 CX 2 C), with two aromatic residues often flanking the last cysteine. This cysteine cluster usually is located near the N terminus and some 30 -50 residues upstream of the putative SAM-binding site (21). In the corresponding cluster of AtsB we mutated two cysteines, Cys 39 and Cys 42 , to alanines. We also mutated three further conserved cysteines that have been predicted to coordinate two additional FeS centers, Cys 270 on the one hand and Cys 331 and Cys 334 on the other (25). All five mutants could be expressed in E. coli at levels comparable with that of wild type AtsB, and the proteins were largely soluble (Fig. 5A). When measuring the activity of co-expressed AtsA, it was found that all five mutants were fully impaired in activating AtsA. After MALDI-TOF MS analysis of AtsA for FGly modification it turned out that in all five cases not even trace amounts of FGly were detectable (Fig. 5A).
In vitro interaction experiments showed that only AtsB mutants C39A and C42A were able to associate with GST-AtsA, albeit with reduced efficiency (55 and 16% of wild type AtsB control, respectively; Fig. 5B). The other mutants showed no interaction. Binding of [ 3 H]SAM also was markedly reduced. Only AtsB-C39A still bound residual amounts of [ 3 H]SAM (ϳ10% of control; Fig. 5B). For all other mutants radioactivity in the GSH eluate was in the background range. Three of the five mutants also were analyzed for their capability of in vitro FGly formation. In line with the in vivo data, none of the cysteine mutants, including AtsB-C39A and AtsB-C42A, was able to generate FGly in the AtsA substrate (Fig. 5C). In conclusion, all three predicted FeS centers seem to be required for FGly modification. Mutation of the second and third cysteine cluster abolishes in vitro binding of SAM and the AtsA substrate, whereas mutation of the first cluster allows for some residual binding (Fig. 5B). This residual binding is blocked in the presence of SAHC (Fig. 5C), suggesting that binding to GST-AtsA is SAM-dependent also for AtsB-C39A. The ternary AtsB⅐SAM⅐AtsA complex of this mutant, however, shows no turnover, indicating that AtsB-C39A is catalytically inactive.
Differential effects on AtsA-AtsB interaction and FGly-generating activity were also observed when adding iron chelators such as EDTA or o-phenanthroline to the in vitro reaction. Whereas EDTA but not o-phenanthroline blocked AtsA-AtsB association, both compounds fully impaired FGly modification (Fig. 5D). These results corroborate the conclusion drawn from the cysteine mutants of AtsB, namely that AtsB carries FeS centers that are essential for its enzymatic function as a radical SAM protein.

DISCUSSION
AtsB Acts as a Radical SAM Protein-Here we show that AtsB acts as a radical SAM protein involving SAM and at least one FeS center as critical co-factors. The essential requirement for SAM was shown in vivo, after mutagenesis of the putative SAM-binding site (see below) and also in in vitro experiments. Here binding of AtsB to AtsA was stimulated by SAM in a concentration-dependent manner, was impaired by its demethylated analog SAHC, and was also dependent on an intact SAM-binding site. In fact, 3 H-labeled SAM was recruited into a ternary AtsB⅐SAM⅐AtsA complex in a roughly 1:1:1 ratio (Fig.  2). Most significantly, we could demonstrate in vitro FGly generation in the GST-AtsA substrate mediated by freshly prepared AtsB in dependence of SAM. This in vitro modification was furthermore dependent on temperature and incubation time and required an intact serine-type modification motif in GST-AtsA ( Fig. 3 and data not shown). It should be noted that already the formation of a stable AtsB⅐SAM⅐GST-AtsA complex was a slow and strictly temperature-dependent process (Fig. 1, B and C). This suggests that recognition of the linear modification motif by AtsB is rate-limiting under the applied in vitro conditions and may require unfolding of the AtsA (residues 21-112) appendix of the GST domain. In fact, elongation of the AtsA appendix at its N or C terminus interfered with in vitro interaction (not shown). In vivo, chaperones or components of the early secretory pathway may maintain the modification competence of newly synthesized AtsA polypeptides. Also our earlier observation that the signal peptide of AtsA, targeting the protein for export to the periplasm, stimulates FGly modification 50 -100-fold, supports the idea that an export-competent, largely unfolded conformation is required for efficient FGly modification (19).
SAM-binding Site of AtsB-By site-specific mutagenesis we found that the sequence 83 GGE 85 is crucial for AtsB function. Substitution of each of these three residues by alanine drastically impaired FGly modification activity in vivo (Fig. 4A). In vitro, none of the mutants was able to bind SAM, to interact with AtsA, or to generate FGly (Fig. 4, B and C). Mutation of the juxtaposed Pro 86 impaired formation of the ternary AtsB⅐SAM⅐AtsA complex partially. GGEP perfectly matches a semi-consensus sequence in an alignment block of the radical SAM superfamily (21). Another member of this huge family is the HemN protein of E. coli, which catalyzes oxygen-independent oxidative decarboxylation of coproporphyrinogen-III.
HemN carries the corresponding GGTP sequence at positions 112-115. A double mutant in which Gly 113 and, in addition, Gly 111 that directly precedes GGTP, both were substituted by valine no longer showed enzymatic activity (31). The binding of SAM by wild type HemN and its inactivation in the mutant, however, was not assayed. The crystal structure of HemN, which is the first three-dimensional structure of a radical SAM protein, was solved recently (32). It verified that Gly 112 and Gly 113 make contact to two SAM molecules bound in the active site tunnel close to the FeS center (see below). The requirement of two SAM molecules may be specific for HemN, which catalyzes two consecutive oxidative decarboxylation reactions on The presence (ϩ) or absence (Ϫ) of FGly in AtsA is given, which was detected as DNP-hydrazone derivative by MALDI-TOF MS analysis of its tryptic peptides. B, in vitro interaction reactions of GST-AtsA and AtsB, expressed in wild type, C39A, C42A, C270A, C331A, or C334A mutant form, were incubated and analyzed as described for Fig. 4B. The positions of the detected proteins, the relative amounts of AtsB, and the amount of SAM-associated radioactivity recovered in a 33% aliquot of the GSH eluate are given. C and D, for in vitro FGly modification, GST-AtsA and AtsB expressed in wild type, C39A, C42A, or C270A mutant form were incubated and analyzed as described for Fig. 4C. The presence (ϩ) or absence (Ϫ) of FGly in GST-AtsA is given, which was detected as DNP-hydrazone derivative by MALDI-TOF MS analysis of its tryptic peptides. In D, 500 M SAM and, where indicated, 2 mM EDTA or 2 mM o-phenanthroline were present during the in vitro reaction. the same substrate (see Ref. 32).
Interestingly, Shields et al. (33) located 98 GXG 100 and 180 EE 181 motifs in phosphatidylethanolamine N-methyltransferase that seem to cooperate in SAM binding. These separate sequence elements were found in subgroups of SAM-utilizing enzymes. From the crystal structure of several SAM-dependent methyltransferases, it is known that a glycine-rich motif (motif I) and a separate motif (post I) containing a negatively charged amino acid make direct contact to SAM (30). Also the HemN structure revealed that the negative charges of Asp 209 and Glu 145 , the latter in direct proximity to Gly 112 , make contact to ribose hydroxyls of the two bound SAM molecules (32). The linear GGE motif of AtsB includes both the glycines and a negatively charged residue that, as shown here, are critical for binding of SAM.
Mechanism of AtsB Function-Like other radical SAM proteins, AtsB needs a specific FeS center to facilitate reductive cleavage of SAM. Apart from the correct position, 30 -50 residues upstream of the SAM-binding site, and perfect match of the 35 CX 3 CXYCY 43 cysteine cluster of AtsB with the corresponding semiconsensus sequence of radical SAM proteins (21), we could provide experimental evidence for the critical role of this cluster. Mutagenesis of the corresponding cysteines fully inactivated the FGly-generating activity of AtsB in vivo and in vitro (Fig. 5, A-C). Interestingly, the same result was obtained after mutagenesis of two additional cysteine clusters, suggesting the existence of two further important FeS centers in the C-terminal half of AtsB, which are not shared with radical SAM proteins in general. Direct evidence for the participation of metal ions came from the observation that chelating agents such as EDTA or o-phenanthroline completely blocked the FGly modification activity of AtsB (Fig. 5D).
The specific FeS center of radical SAM proteins in its [4Fe-4S] ϩ form is active for reductive cleavage of SAM (23)(24). Because it is coordinated by only three cysteines, one iron coordination position is vacant, thereby explaining its oxygen sensitivity (22,32). On the other hand, the unique iron, which is not coordinated by a cysteine, allows for SAM binding through association of the SAM sulfonium directly with this iron and/or with an adjacent sulfur atom in the [4Fe-4S] center and a bidentate chelation of the iron by the methionyl carboxylate and amine (32,34,35). These close interactions facilitate transfer of an electron from the FeS center to SAM, resulting in its reductive cleavage to form methionine and the highly oxidizing deoxyadenosyl radical. The latter can directly abstract a hydrogen atom from unactivated C-H bonds. In case of AtsB a hydrogen would be abstracted from C␤H 2 -OH of Ser 72 in AtsA, leading to production of deoxyadenosine and the substrate radical [CH-OH] ⅐ . The latter has to be further oxidized, possibly by transferring an electron to the oxidized [4Fe-4S] 2ϩ center, thereby producing FGly (CHOϩH ϩ ) and regenerating the reduced [4Fe-4S] ϩ center.
To avoid formation of the highly reactive deoxyadenosyl radical in the absence of substrate, the generation of the reduced [4Fe-4S] ϩ form and/or SAM binding have to be tightly regulated. The former can be achieved by controlled reduction through other FeS centers present in AtsB, and the latter can be achieved by cooperativity of substrate (AtsA) and SAM binding. For biotin synthase, a well studied radical SAM protein, Ugulava et al. (36) observed that both the rate of binding and the affinity for SAM increased significantly in the presence of substrate and furthermore that substrate bound only in the presence of SAM. In our in vitro system we noticed a pronounced stimulation of AtsB binding at low, nonsaturating SAM concentrations (Fig. 2B), which may indicate such cooperativity.
In view of the model outlined above, the critical [4Fe-4S] ϩ/2ϩ center with its unique iron contributes to SAM binding. We observed that mutating the corresponding cysteine cluster (Cys 39 and Cys 42 ) led to catalytically inactive AtsB protein in which SAM binding was much stronger affected than association with AtsA (Fig. 5B). This may reflect the contribution of the [4Fe-4S] center to SAM binding.
Further insight into molecular aspects of AtsB function most probably will require analysis under anaerobic conditions and to work with AtsB protein (wild type and mutant) that is characterized with respect to its iron content. Studies of its FeS center(s) by EPR spectroscopy in the absence and presence of SAM and a substrate peptide should clarify basic properties of those centers and their interconversions. For understanding the mechanism of the radical SAM family in general, consisting of an enormous variety of enzymes, the first three-dimensional structure that is now available (32) will be extremely valuable. The structures of other family members will reveal common principles and also the variations underlying radical SAMmediated catalysis.