Structural Insights into Substrate Recognition and Catalysis in Outer Membrane Protein B (OmpB) by Protein-lysine Methyltransferases from Rickettsia*

Rickettsia belong to a family of Gram-negative obligate intracellular infectious bacteria that are the causative agents of typhus and spotted fever. Outer membrane protein B (OmpB) occurs in all rickettsial species, serves as a protective envelope, mediates host cell adhesion and invasion, and is a major immunodominant antigen. OmpBs from virulent strains contain multiple trimethylated lysine residues, whereas the avirulent strain contains mainly monomethyllysine. Two protein-lysine methyltransferases (PKMTs) that catalyze methylation of recombinant OmpB at multiple sites with varying sequences have been identified and overexpressed. PKMT1 catalyzes predominantly monomethylation, whereas PKMT2 catalyzes mainly trimethylation. Rickettsial PKMT1 and PKMT2 are unusual in that their primary substrate appears to be limited to OmpB, and both are capable of methylating multiple lysyl residues with broad sequence specificity. Here we report the crystal structures of PKMT1 from Rickettsia prowazekii and PKMT2 from Rickettsia typhi, both the apo form and in complex with its cofactor S-adenosylmethionine or S-adenosylhomocysteine. The structure of PKMT1 in complex with S-adenosylhomocysteine is solved to a resolution of 1.9 Å. Both enzymes are dimeric with each monomer containing an S-adenosylmethionine binding domain with a core Rossmann fold, a dimerization domain, a middle domain, a C-terminal domain, and a centrally located open cavity. Based on the crystal structures, residues involved in catalysis, cofactor binding, and substrate interactions were examined using site-directed mutagenesis followed by steady state kinetic analysis to ascertain their catalytic functions in solution. Together, our data reveal new structural and mechanistic insights into how rickettsial methyltransferases catalyze OmpB methylation.

Rickettsia belong to a family of Gram-negative obligate intracellular infectious bacteria that are the causative agents of typhus and spotted fever. Outer membrane protein B (OmpB) occurs in all rickettsial species, serves as a protective envelope, mediates host cell adhesion and invasion, and is a major immunodominant antigen. OmpBs from virulent strains contain multiple trimethylated lysine residues, whereas the avirulent strain contains mainly monomethyllysine. Two protein-lysine methyltransferases (PKMTs) that catalyze methylation of recombinant OmpB at multiple sites with varying sequences have been identified and overexpressed. PKMT1 catalyzes predominantly monomethylation, whereas PKMT2 catalyzes mainly trimethylation. Rickettsial PKMT1 and PKMT2 are unusual in that their primary substrate appears to be limited to OmpB, and both are capable of methylating multiple lysyl residues with broad sequence specificity. Here we report the crystal structures of PKMT1 from Rickettsia prowazekii and PKMT2 from Rickettsia typhi, both the apo form and in complex with its cofactor S-adenosylmethionine or S-adenosylhomocysteine. The structure of PKMT1 in complex with S-adenosylhomocysteine is solved to a resolution of 1.9 Å. Both enzymes are dimeric with each monomer containing an S-adenosylmethionine binding domain with a core Rossmann fold, a dimerization domain, a middle domain, a C-terminal domain, and a centrally located open cavity. Based on the crystal structures, residues involved in catalysis, cofactor binding, and substrate interactions were examined using sitedirected mutagenesis followed by steady state kinetic analysis to ascertain their catalytic functions in solution. Together, our data reveal new structural and mechanistic insights into how rickettsial methyltransferases catalyze OmpB methylation.
Methylation of outer membrane proteins has been implicated in rickettsial virulence. Rickettsia belong to a family of obligatory intracellular infectious bacteria that are the causative agents of typhus and spotted fever (1). The bacterial outer membrane of Rickettsia contains a major surface protein called OmpB 5 that occurs in all rickettsial species and accounts for up to 15% of total cellular proteins (2)(3)(4). OmpB has been shown to mediate host cell adhesion, attachment, and invasion (5)(6)(7). OmpB belongs to the family of autotransporters (8) and is a major immunodominant antigen (3). The precursor of OmpB consists of a signal peptide, a passenger domain, and a C-terminal ␤-barrel domain (9). The passenger domain of OmpB has been shown to undergo methylation at its lysine residues, and this methylation appears to be associated with rickettsial pathogenicity and immunogenic response (10 -15). OmpB was first found to be methylated by amino acid composition analysis and later confirmed using mass spectrometric methods. The levels of methylation of OmpB from several virulent and avirulent strains appear to correlate well with the virulence level of the strains. In-depth analysis of methylation profiles, using semiquantitative integrated LC-MS/MS methods on the locations, states, and levels of methylated lysine residues in OmpB purified from several rickettsial strains, revealed that (i) OmpBs from virulent strains contain clusters of highly trimethylated lysine residues and (ii) OmpB from the avirulent strain contains primarily monomethyllysine residues and no trimethyllysine (16).
Posttranslational protein methylation has been shown to play a major role in regulating biological processes (17,18) ranging from chemotaxis to epigenetics (19 -21), transcription regulation (22), translation (23), and cell signaling (24). If methylation of OmpB indeed causes virulence, rickettsial virulence could be mediated in part by cellular proteins involved in the regulation of these processes.
Two classes of rickettsial protein-lysine methyltransferases (PKMTs) have been identified and termed PKMT1 and PKMT2 via bioinformatics analysis, cloning, overexpression, and char-acterization in terms of their enzyme activity using recombinant OmpB fragments as the substrates (25). Both classes of PKMTs catalyze methylation of recombinant OmpB but show distinct levels of specificity and types of methylation. At distinct sites, PKMT1 catalyzes primarily monomethylation and produces substantially lower levels of di-and trimethylated lysine, whereas PKMT2 catalyzes almost exclusively trimethylation at relatively specific sites (16). In recombinant fragments of OmpB, RpPKMT1 can catalyze monomethylation of its 39 lysyl residues, and 14 of these residues can be trimethylated by RpPKMT2 (16). Homologous PKMTs have been found in more than 40 different rickettsial species including RP789 from Rickettsia prowazekii (RpPKMT1), RT776 from Rickettsia typhi (RtPKMT1), RP027-028 from R. prowazekii (RpPKMT2), and RT0101 from R. typhi (RtPKMT2). The locations, states, and levels of methylation in methylated recombinant OmpB fragments using these PKMTs and the methyl donor AdoMet were determined using LC-MS/MS. The observed lysine methylation profiles from in vitro methylation correlate well with those found in native OmpBs purified directly from Rickettsia (16).
The results indicate that virulent strains contain and express both PKMT1 and PKMT2, but the avirulent strain contains a frameshift mutation of PKMT2 gene and does not possess active PKMT2 (13). The primary substrate of rickettsial PKMTs appear to be limited to OmpB because these purified PKMTs fail to catalyze the methylation of histones or Escherichia coli proteins (25). Furthermore, PKMT1 and PKMT2 exhibit unusually broad amino acid sequence specificity and produce distinct methylation profiles in OmpB. To our knowledge, with the exception of PrmA methyltransferase that catalyzes trimethylation at multiple residues of ribosomal protein L11 (26) and of archaeal protein-lysine methyltransferase that multimethylates Cren7, a chromatin protein (27), no PKMT that catalyzes methylation at multiple sites of a protein in a manner similar to that exhibited by rickettsial PKMTs has been reported. Structural comparison of PKMT1 and PKMT2 may uncover insights into their different product specificity. Comparative structural analyses with known structures of PKMTs (28 -30) could reveal the physical and chemical bases that underlie the unusual catalytic properties of rickettsial PKMTs.
In this study, we determined the crystal structures of RpPKMT1 and RtPKMT2 as well as the complexes with the cofactor, AdoMet or AdoHcy. Our structural data provide the first look of this family of PKMTs that specifically target OmpB, allowing us to identify conserved active site residues and a putative substrate binding cleft, and revealed their catalytic roles using site-directed mutagenesis and enzyme kinetic analyses. Together, our data provide structural and mechanistic insights into rickettsial PKMTs and enable us to propose a model that describes the molecular basis for how the rickettsial PKMTs interact with AdoMet and OmpB for substrate specificity and catalysis.

Results
The RpPKMT1 and RtPKMT2 Crystal Structures-Crystal structural analysis was performed to investigate the catalytic action of RpPKMT1 and RtPKMT2 in their methylation of OmpB. Fig. 1 shows the amino acid sequence alignment of RpPKMT1 and RtPKMT2. The sequence of RpPKMT1 exhibits a 44% identity with that of RtPKMT2 and an N-terminal extension of 28 residues. RpPKMT1 catalyzes predominantly lysine monomethylation, whereas RtPKMT2 catalyzes almost exclusively trimethylation (16,25). To determine the crystal structure, the purified recombinant RpPKMT1, SeMet-labeled RpPKMT1, and RtPKMT2 were prepared and crystallized. All structures were solved in space group P2 1 . Selenium single wavelength anomalous dispersion phasing was used to solve the structures of SeMet-substituted RpPKMT1 to 2.9-Å resolution. This structure was then used as a search model to solve the native structures of RpPKMT1 and RtPKMT2 to a final resolution of 2.6 and 3.1 Å, respectively ( Table 1). The RpPKMT1 structure consists of modeled residues Tyr 43 -Val 553 of chain A and chain B, whereas the RtPKMT2 structure consists of modeled residues Pro 14 -Gly 534 of chain A and chain B. For all structures, two molecules were found in the asymmetric unit related by 2-fold symmetry with significant interactions, indicating a possible homodimer.
The crystal structures of RpPKMT1 and RtPKMT2 are shown in Fig. 2. The overall three-dimensional folds of RpPKMT1 and RtPKMT2 are well conserved with a root mean square deviation of 1.51 Å (Fig. 2 RpPKMT1 and RtPKMT2 share an open cleft located along the base of the AdoMet binding domain that may serve as the putative protein substrate binding site. A number of loops from each of the adjacent domains are ideally positioned to interact with protein substrates. The most prominent structural difference between RpPKMT1 and RtPKMT2 occurs in the C-terminal domain where RtPKMT2 contains an elongated loop (His 440 -Ser 451 ) (Fig. 2, C and D). This extension is mediated by an insertion of six amino acid residues (Asn 444 -Met 449 ) in RtPKMT2 as indicated in the sequence alignment in Fig. 1. Interestingly, this loop is located in close proximity to the AdoMet binding domain and may play a role in enzyme-substrate interactions.
The RpPKMT1 and RtPKMT2 crystal structures are arranged as a 2-fold symmetric dimer mediated by the mostly helical dimerization domain (Fig. 2E). To determine whether the dimers observed in the crystal structures were the biological unit, size exclusion chromatography with multiangle light scattering was performed. The data show that both PKMTs also exist as dimers in solution (Fig. 2F). The buried surface area of the dimers is 1540.6 and 1451.3 Å 2 for RpPKMT1 and RtPKMT2, respectively, with further stabilization from multiple hydrogen bonds along the dimer interface. The overall electrostatics of the two structures differ in various regions and may contribute to interactions with substrate; however, further work is needed to confirm this hypothesis (Fig. 2G).
Crystal Structures of RpPKMT1 in Complex with AdoMet or AdoHcy-The crystal structures of RpPKMT1 bound with AdoMet or AdoHcy were solved by crystal soaking and co-crystallization experiments, respectively. Both ligands were found to bind within the AdoMet binding domain along the conserved Rossmann fold (Fig. 3A). The complex with AdoMet yielded a lower resolution structure with the AdoMet ligand showing disorder at the binding site as evidenced by poor density along the carboxyl end of the AdoMet molecule (Fig. 3B). However, the complex with AdoHcy was solved at 1.9-Å resolution with significantly more order in the ligand (Fig. 3C) and surrounding residues. Fig. 3D shows the network of amino acid residues in RpPKMT1 that interact with AdoHcy.
Both ligands were found in extended conformations within the deep narrow elongated pocket formed along the base of the AdoMet binding domain (Fig. 3A). The methyl donor side of AdoHcy faces the open cleft, whereas the interior face interacts with the AdoMet binding domain. The structures of AdoHcy and AdoMet superimpose well, and therefore they both bind at essentially the same site in RpPKMT1. Furthermore, structural comparison of apo-RpPKMT1 with the AdoMet-and AdoHcybound complexes reveals few overall structural changes upon binding of the cofactor. Slight conformational changes in the side chains of Ser 129 and Leu 103 are observed moving toward the ligand as well as some slight changes along the dimerization FIGURE 1. Amino acid sequence alignment of RpPKMT1, RpPKMT2, RtPKMT1, and RtPKMT2. Amino acid sequence alignment was obtained using ClusterW and ESPript (49). Identical residues are highlighted in blue, and conserved residues are shown in red. The secondary structure shown above the sequences is based on the crystal structure of the RpPKMT1-AdoHcy complex. The structural domains consist of the AdoMet binding domain (green), the dimerization domain (yellow), the middle domain (blue), and the C-terminal domain (magenta).
interface. Overall no large domain shifts or conformational changes were observed between the apo and holo structures. The same results as RpPKMT1 were observed for RtPKMT2 in complex with AdoHcy.
Catalytic and Structural Effects Due to Elongated Loop Insertion into RpPKMT1 and Deletion from RtPKMT2-Structural analysis of RpPKMT1 and RtPKMT2 revealed that the six-residue insertion, Asn 444 -Met 449 , presents as an elongated loop in the C-terminal domain of RtPKMT2 (Figs. 1 and 2D). The location of the elongated loop in RtPKMT2 and the amino acid residues in the stick model are shown in Fig. 4, A and B, respectively. This loop is ideally positioned along the AdoMet binding site and can also potentially interact with the protein substrate ( Fig. 4C). Because RpPKMT1 functions primarily as a monomethyltransferase and RtPKMT2 mainly catalyzes the trimethylation reaction, we hypothesized that this loop may account for the different methylation reactions catalyzed by the two PKMTs. Therefore, we explored the catalytic function of this loop first by deleting residues Asn 444 -Met 449 in RtPKMT2 and second by inserting the Asn 444 -Met 449 loop of RtPKMT2 into the corresponding location in RpPKMT1. Inserting the six-residue loop into RpPKMT1 showed a faster initial increase for the methylation process relative to that observed with the WT enzyme under the indicated experimental conditions (Fig. 5A). Fig. 5B shows that deletion of residues Asn 444 -Met 449 in RtPKMT2 led to an increase in the initial rate for the incorporation of the radioactive methyl group into the recombinant OmpB(AN) from [methyl-3 H]AdoMet. The apparent catalytic and Michaelis-Menten constants of RpPKMT1, the loop-inserted RpPKMT1, RtPKMT2, and the loop-deleted RtPKMT2 were analyzed using much lower enzyme concentrations, and the results are shown in Table 2. These data reveal that loop insertion into RpPKMT1 did not alter much the values of k cat , whereas the value of K m was noticeably reduced. When the elongated loop was deleted from RtPKMT2, the values of both k cat and K m were elevated. Table 2 also shows that the catalytic efficiency (k cat /K m ) for the loop-containing methyltransferases is about 2-fold higher than that observed in the absence of the elongated loop. Note that the comparative analysis shown in Fig. 5, A and B, and the observed values of k cat shown in Table 2 indicate that the rate of the RpPKMT1-catalyzed monomethylation reaction is about 200-fold faster than that of RtPKMT2catalyzed trimethylation reaction.
To resolve whether the increase in the rate of methylation observed with the loop-deleted RtPKMT2 is attributed to the possibility that deletion of the elongated loop may partially convert the trimethyltransferase to catalyze the formation of monomethylated OmpB, we analyzed the methylation profiles of OmpB using LC-MS/MS methods (16). Fig. 5C shows that the loop insertion to RpPKMT1 did not elevate the production of trimethyllysine residues. Similarly the loop deletion of RtPKMT2 also did not convert the enzyme from a tri-to a monomethyltransferase (Fig. 5D). In essence, the elongated loop does not play an important role in switching the enzyme from a mono-to a trimethyltransferase or vice versa.
where a test set, T (5% of the data), is omitted from the refinement. e Performed using MolProbity. Circular dichroism and intrinsic fluorescence were used to monitor whether the loop mutations could induce significant conformational changes of PKMTs. Fig. 6A shows that there are no appreciable CD spectral changes caused by loop deletion from RtPKMT2 or insertion into RpPKMT1. However, notable changes of the intensity and the emission maximum wavelength of the intrinsic fluorescence of RtPKMT2 (Fig. 6B) were detected due to loop deletion, suggesting that the observed changes in k cat and K m of the loop-deleted RtPKMT2 could be attributed to the localized perturbation of conformation induced by the removal of the loop.
Catalytically Essential Residues at the AdoMet Binding Site-Crystal structural analysis revealed that the cofactor AdoMet/ AdoHcy forms a network of interactions with RpPKMT1 (Fig.  3D). They include hydrogen bonds with residues Tyr 48 (OH to carboxyl moiety of Hcy), Asp 102 (␤-carboxyl to cis-diol of ribose), His 146 (backbone carbonyl to ␣-NH 2 of Hcy), Gly 79 (backbone carbonyl to ␣-NH 2 of Hcy), Ser 129 (OH to N4 of adenine), and Ile 130 (backbone nitrogen to N3 of adenine). Side chains of Ile 130 and Leu 103 stack from opposite sides of the adenine ring. The catalytic roles of these residues at the AdoMet binding site in RpPKMT1 were probed by mutating each of these residues to Ala. The single point mutants Y48A, E77A, N85A, D102A, and H146A all exhibited no PKMT activity. The loss of PKMT activity in D102A and H146A mutants is likely due to the loss of critical hydrogen bonds between RpPKMT1 and AdoMet (Fig. 3D). The apparent K m for AdoMet and OmpB(AN) and apparent k cat were determined for those mutants that retained their PKMT activity including Y48F, L103A, I130A, and C145A. As shown in Table 3, these mutations affect the values of both K m and k cat . Interestingly, the mutant Y48F exhibited a significant decrease in the k cat value and increase in K m for AdoMet, revealing the effect of eliminating the hydrogen bond formation between OH of the tyrosine residue and the carboxyl moiety of AdoMet. In addition, substituting Ile 130 with Ala also yielded an increase in K m AdoMet and a moderate decrease in k cat . Together, these observations indicate that the amino acid residues located at the AdoMet binding site of RpPKMT1 are involved in both the binding of AdoMet and the catalytic action of RpPKMT1. Similar to RpPKMT1, the crystal structure of RtPKMT2 reveals that Asn 57 , Asp 74 , Leu 75 , Ile 102 , and His 118 are located at the cofactor binding site found in the RpPKMT1-AdoHcy complex (Fig. 4C), suggesting that these residues may interact with AdoMet during catalysis. Consistent with this notion, alanine mutation of these residues led to drastic changes in the activity of RtPKMT2. Our experiment revealed that the RtPKMT2 point mutants D74A and H118A exhibited no PKMT activity (results not shown), whereas the mutants L75A and I102A remained active (Table 3). Steady state kinetics revealed that both K m and k cat of the active mutants were altered significantly from those found with the wild type RtPKMT2. With the exception of the relatively inactive RpPKMT1(Y48F), the effects on K m caused by point mutations of the residues at the AdoMet binding domain were comparable for RpPKMT1 and RtPKMT2. However, the value of k cat decreases with the RpPKMT1 mutants but increases with RtPKMT2 mutants. Clearly, mutation studies show that these residues play critical roles in the methylation catalysis in accord with their interactions with the cofactor at the AdoMet binding site found in the crystal structures.
Identification of the Putative Protein Substrate Binding Site in RtPKMT2-To identify those residues that are important for protein substrate binding, we first aligned the amino acid sequences of four PKMTs (RpPKMT1, RpPKMT2, RtPKMT1, and RtPKMT2), which are known to methylate OmpB (Fig. 1). We then mapped all fully conserved residues to the surface of the RtPKMT2 crystal structure with the rationale that those residues that are important for binding OmpB would be fully conserved in all four PKMTs. Based on this analysis, we were able to identify a number of solvent-exposed residues in prox-imity to the AdoMet binding site that we hypothesized may serve as the substrate binding site. Among them, residues in the elongated loop of RtPKMT2 that could potentially be involved in OmpB interaction are shown in Fig. 7A. In addition, crystal structural analysis also shows that residues Tyr 219 , Trp 123 , Trp 156 , Phe 291 , and Tyr 224 form a hydrophobic pocket facing the methyl group of AdoMet. This hydrophobic environment is expected to lower the local pK a and to enhance the nucleophilicity of the amino group to facilitate the methylation reaction.
The potential catalytic function of the amino acid residues in RtPKMT2 was examined by monitoring the change in its PKMT activity by site-directed mutagenesis ( Fig. 7B and Table  4). Mutations of Trp 123 , Ser 149 , Asn 151 , Arg 163 , and Phe 291 to alanine eliminated their PKMT activity detectable under standard assay conditions, whereas mutants W156A, Y219A, and E208A retained low MT activity. These eight residues are located along the flat base of the AdoMet binding domain (Fig.  7B). The observed inactivation and inhibition caused by mutation to Ala of these residues, which are distant to the methyl donor, imply that they interact with OmpB during the catalytic action of PKMT2.
To further investigate the mechanism by which site-directed mutagenesis alters catalytic activity, the K m for OmpB(AN) and k cat of the highly active mutant L211A and the moderately active Y219A mutant were determined (Table 4). In addition, residues Tyr 342 , Tyr 343 , and Glu 348 in the middle domain were also mutated to Ala to probe the possible involvement of the middle domain during the catalytic action of RtPKMT2. Our results (Table 4) show that mutants Y219A and Y343A induced a significant and specific reduction in the catalytic rate con- stant, whereas mutants L211A, E348A, and Y342A altered the values for both K m for OmpB(AN) and k cat .
Taken together, the catalytic effects from alanine scanning of amino acid residues along the base of the AdoMet binding domain (Figs. 4C and 7B), in the middle domain (depicted in Table 4), and in the elongated loop of RtPKMT2 (Fig. 4B) (shown in Table 2) imply that each of the mutated residues has the potential to directly interact with the enzyme-bound OmpB or that the mutated residues could induce an allosteric effect to alter OmpB binding affinity and/or the catalytic activity.

Discussion
Here we present the first crystal structures of PKMTs known to target specifically an OMP in Gram-negative bacteria. Our structural studies reveal a number of features shared between RpPKMT1 and RtPKMT2 and likely with other members of this family. This family has a unique overall fold compared with other families of MTs, albeit the AdoMet binding domain is well conserved. PKMT1 and PKMT2 share four structural domains consisting of a core AdoMet binding domain, a dimerization domain, a middle domain, and a C-terminal domain. In comparison with other known structures of MTs in the Protein Data Bank, PKMT1 and PKMT2 possess a Ross-mann fold characteristic of type 1 MTs, whereas most structures of protein-lysine MTs in the Protein Data Bank have a SET domain in their active site with the exception of PrmA (26) and the DOT1 family (29,31,32). Unlike known PKMTs, PKMT1 and PKMT2 contain the middle and C-terminal domains. Furthermore, both enzymes are dimers, and this is likely a shared feature of this family.
RpPKMT1 catalyzes primarily monomethylation, whereas RtPKMT2 catalyzes trimethylation (16), commonly known as product specificity. Our study reveals that the three-dimensional structures of these enzymes are highly conserved. Given the structural similarity between the two PKMTs, it is intriguing that they exhibit such distinct product specificity.
One major structural difference between RpPKMT1 and RtPKMT2 is the presence of a six-residue insertion (Asn 444 -Met 449 ) in a C-terminal loop in RtPKMT2 that projects across the top of the open cleft as an elongated loop over where the AdoMet binding site is located. In our attempt to investigate whether this elongated loop of RtPKMT2 may participate in mediating the trimethyltransferase activity of RtPKMT2, we analyzed the methylation profiles of OmpB catalyzed by RtPKMT2, loop-deleted RtPKMT2, RpPKMT1, and loop-inserted RpPKMT1. The results (see Fig. 5, C and D) reveal that  Table 2. C, mutated residues in the AdoMet binding site are shown in the stick model (RpPKMT1 in green and RtPKMT2 in gray). The resulting effects of mutants that retained MT activity on the methylation of OmpB(AN) are shown in Table 3.
this elongated loop does not play an important role in switching the enzyme from a mono-to a trimethyltransferase or vice versa.
RpPKMT1 contains an additional 28 N-terminal residues in the sequence alignment with RtPKMT2 (Fig. 1). Deletion of the N-terminal residues from RpPKMT1 (RP789) or RtPKMT1 (RT776) does not inactivate its PKMT activity and causes a limited elevation of di-and trimethylation as monitored using the LC-MS/MS method (16). This observation indicates that the N-terminal extension in PKMT1 partially modulates the formation of multimethylated OmpB. It should be pointed out that the crystal structures for both RpPKMT1 and RtPKMT2 reveal that their N-terminal sequences are disordered and provide no structural basis for altering product specificity.
In addition to the elongated loop in RtPKMT2 and the N-terminal extension in RpPKMT1, it is intriguing to note that the two PKMTs also differ in their substitutions of five conserved Tyr residues in PKMT1 to Phe in PKMT2. Mechanistic studies of the SET domain in histone MTs using site-directed mutagenesis coupled with crystal structural analysis reveal that substituting a Tyr residue at the active site with a Phe attenuates hydrogen bonding to a structurally conserved water molecule adjacent to the Phe/Tyr switch and facilitates its dissociation. When this water molecule dissociates, it enables the side chain of the monomethyllysyl residue to adopt a catalytically competent conformation and space to accommodate the formation of di-or trimethyllysine (33,34). The present crystal structures of PKMTs reveal that three of the five substituted Tyr residues, Tyr 46 , Tyr 48 , and Tyr 175 , are located in close proximity to bound AdoMet in PKMT1. Thus, the crystal structures from the present study advance the likelihood that these Tyr and Phe residues may modulate their catalyzed reactions toward monoand trimethylation by PKMT1 and PKMT2, respectively.
Based on the sequence alignments of PKMTs that target OmpB, we identified the fully conserved residues that may mediate substrate binding. We then mapped these residues to the surface of the RtPKMT2 structure and used mutagenesis studies to identify a putative substrate binding site that is located in close proximity along the AdoMet binding site (Fig.  7A). The AdoMet binding site is within the open cleft. Interestingly, the Ala mutations at the AdoMet binding site Leu 103 and Ile 130 affect the K m value for OmpB(AN) ( Table 3). In addition, RtPKMT2 mutants such as W123A, S149A, N151A, R163A, Y219A, and F291A (shown in Fig. 7B) drastically reduce their catalytic activity for OmpB(AN) methylation. Thus, mutation of conserved residues along this putative OmpB binding site would inactivate RtPKMT2. Together, these results lead us to propose the working model for RtPKMT2 depicted in Fig. 7C. It shows the AdoMet binding site and an OmpB fragment threaded along its putative binding site. A fully extended substrate fragment peptide is positioned along this binding site to illustrate how OmpB may bind into the cleft. Crystallization attempts to capture a complex with a peptide-mimetic substrate have so far been unsuccessful. Fig. 7C is our working model for PKMT2 showing that the peptide substrate is present in an extended conformation that binds to the cleft in PKMT2.
Note that examples of peptides in extended conformation binding to proteins have been reported in the literature, e.g. binding of SNAP-25 to the light chain of botulinum neurotoxin (35), complex formation between the octapeptide repeat of prion protein and the Fab fragment of the POM2 antibody (36), peptide substrate at the binding cleft of Hsp70 chaperone DnaK (36,37), and a number of histone peptides in complex with histone-lysine methyltransferases (34,38). It is conceivable that PKMT binding to an extended conformation of OmpB could facilitate enzymatic methylation at multiple lysyl residues. PKMT1 and PKMT2 target a large number of lysyl residues with diverse amino acid sequences for multisite methylation (16). With the exception of PrmA (26) and archaeal Cren7 (27) PKMTs, most PKMTs with known crystal structures methylate a unique Lys residue located in specific sequence motifs.
PKMTs have an open cleft at bound AdoMet, whereas the SET domain in histone MTs has a well shielded AdoMet and utilizes a narrow channel for the side chain of Lys to relay the ⑀-amino group of the Lys residue being methylated (39 -41). On the one hand, DOT1 histone MT, which is also a type 1 MT, catalyzes specifically the methylation of Lys 79 of histone H3 (29). On the other hand, PrmA shows large domain movements mediated through a flexible linker between the catalytic and substrate binding domains that provides the structural bases for multisite methylation (26). However, OmpB PKMTs exhibit little or no conformational changes upon AdoMet binding. The mecha- nism of the multisite methylation at diverse sequences catalyzed by PKMT1 and RtPKMT2 is not known and remains to be investigated.
In conclusion, rickettsial PKMT1 and PKMT2 catalyze multisite methylation of OmpB. We determined the crystal structures of OmpB monomethyltransferase RpPKMT1 and trimethyltransferase RtPKMT2 and their respective complexes with the cofactor AdoMet/AdoHcy and carried out site-directed mutagenesis and biochemical studies to illuminate substrate binding and catalysis. The structural determination revealed that both PKMT1 and PKMT2 are dimers, contain a sevenstrand Rossmann fold for AdoMet binding, and enfold a cleft. We identified residues that are essential for the cofactor binding and methyl transfer. Mutagenesis studies of PKMTs revealed that the elongated loop does not play an important role in switching the enzyme from a mono-to a trimethyltransferase. This work provides an example of structurally distinct proteins that carry out a common function of AdoMet-depen- dent methyl transfer to lysyl residues in proteins. Together, these results could provide new insights for structure-based design of inhibitors against PKMTs (42,43) and facilitate the development of novel drugs against virulent Rickettsia.

Experimental Procedures
Protein Expression and Purification-The cloning, expression, and purification of RpPKMT1, RtPKMT2, OmpB(AN) (R. typhi OmpB residues 33-744), and OmpB(K) (R. typhi OmpB residues 745-1353) were performed as reported previously (25) with modifications. Briefly, codon-optimized RpPKMT1 and RtPKMT2 gene sequences carrying sequence encoding a tobacco etch virus protease recognition site immediately upstream of the start codon were subcloned at NdeI/ XhoI restriction sites into the pET28a expression vector (Novagen) downstream of the His 6 tag coding sequence, and the plasmid was transformed into E. coli BL21(DE3) cells (Agilent Technologies). The cultures were induced with 0.04 mM isopropyl 1-thio-␤-D-galactopyranoside, and the mixtures were incubated overnight at 22°C with shaking at 250 rpm. Purification was performed using nickel-nitrilotriacetic acid affinity chromatography. The His 6 tag was cleaved by incubating tobacco etch virus-His 6 protease at 4°C overnight, and PKMT was flowed through a second nickel-nitrilotriacetic acid column. The sample was then concentrated and chromatographed on a Sephacryl S100 HR gel filtration column (GE Healthcare) in crystallization buffer (20 mM Tris-HCl, pH 7.5, 200 mM NaCl). Protein in the peak fractions was concentrated to a final concentration of 10 mg/ml, and tris(2-carboxyethyl) phosphine hydrochloride was added to a final concentration of 5 mM. For SeMet-labeled RpPKMT1, B834 competent cells (New England Biolabs) were used along with SelenoMethionine Medium Base plus Nutrient Mix (AthenaES, Baltimore, MD). Protein concentration was determined using a NanoDrop UVvisible spectrophotometer. For the complexes of RpPKMT1 or RtPKMT2 with AdoMet or AdoHcy, PKMT was incubated with 5 mM AdoMet or AdoHcy (Sigma) in crystallization buffer for at least 60 min on ice prior to crystallization.
Data Collection and Structure Determination-Initial crystals were screened using an in-house x-ray diffractometer (Rigaku MicroMax-007 HF microfocus x-ray generator, Raxis IV ϩϩ detector) with final data sets collected at either the General Medicine Sciences and National Cancer Institute Collaborative Access Team (GM/CA-CAT) or Southeast Regional Collaborative Access Team (SER-CAT) beamline at the Advanced Photon Source at Argonne National Laboratory. All data were processed using HKL2000 (44), and statistics are summarized in Table 1. For experimental phasing, SeMet-substituted RpPKMT1 was prepared using B834(DE3) cells (Merck Millipore) grown in SelenoMet medium (Molecular Dimensions) supplemented with SeMet (40 mg/l) and purified using the same protocol as for the native protein sample. The initial RpPKMT1 structure was solved by selenium single wavelength anomalous dispersion (AutoSol) and a starting model automatically built (AutoBuild) and refined using PHENIX (45,46). Molecular replacement was then used to solve the native and complex structures of RpPKMT1 and the RtPKMT2 native structure. A difference map was used to locate the AdoMet and AdoHcy ligands. All model building was performed using Coot (46,47), and final refinement was performed using PHENIX and CCP4 (48). Figures were prepared using PyMOL (Schrödinger), and final editing was performed with Adobe Illustrator. Root mean square deviation analysis was carried out using PyMOL.
Site-directed Mutagenesis-Mutagenesis was carried out using QuikChange Lightning mutagenesis kits (Agilent) according to the manufacturer's protocol. Primers for site-directed mutagenesis were designed using the QuikChange Primer Design online tool (Agilent). Sequences of primers are available upon request. Mutant plasmids were purified from overnight culture using a Qiaprep Spin Miniprep kit (Qiagen), and DNA sequences were verified by sequencing analysis (Genewiz Inc.).
PKMT Activity Assay-PKMT activity was assayed by the incorporation of methyl-3 H into protein substrate Omp-B(AN) using [methyl-3 H]AdoMet as described previously (25).
Steady-state Kinetics Analysis-Initial rates of methylation catalyzed by WT or mutants of RpPKMT1 and of RtPKMT2 were determined from the linear portions of the time courses obtained with the indicated concentration of OmpB(AN) and varying concentrations of AdoMet or at constant AdoMet and varying concentrations of OmpB(AN). The reaction was monitored under standard assay conditions using the PKMT radioactivity assay at 37°C (25). The values of K m and k cat were determined by direct curve fitting using KaleidaGraph (Synergy) and Sigmaplot. Recombinant OmpB(AN) contains multiple lysyl residues that are methylated under the assay conditions. In

TABLE 4 Apparent kinetic constants of Ala mutants of RtPKMT2
The initial rates of methylation catalyzed by RtPKMT2 mutants L211A, Y219A, E348A, Y342A, and Y343A were analyzed using 0.5 M mutant enzymes. The concentrations of OmpB(AN) were varied from 0.5 to 4.0 M at 0. 16  addition, multiple methyl groups can be incorporated into each ⑀-amino moiety of lysyl residues. The values of K m and k cat of various methylation reactions at different lysyl residues may vary in a wide range. Thus, the kinetic constants so obtained are the weighted averages of multiple methylation reactions catalyzed by PKMT and can only be considered as apparent K m and k cat . It should be pointed out that we could not use very high concentrations of OmpB(AN) for our kinetic measurements because OmpB(AN) is a membrane-binding protein, and it tends to form aggregates at high protein concentration in the absence of its membrane counterpart. Mass Spectrometry-Integrated liquid chromatography-tandem mass spectrometry and in-gel digestion with chymotrypsin were performed as described previously (16). Note the LTQ Orbitrap Elite mass spectrometer (Thermo Scientific, San Jose, CA) used in this study routinely yielded a mass error range of 3 ppm or less. This high mass accuracy allowed us to differentiate between trimethylation and acetylation or between dimethylation and formylation in our peptide mass analysis (16). Data analysis and calculation of the normalized fraction were carried out as described (16).
For LC-MS/MS analysis, OmpB(AN) (10 g) and OmpB(K) (5 g) were methylated separately using 10 g of the specified methyltransferase in 50 l of reaction mixtures containing 3.2 mM AdoMet (New England Biolabs) and 8.3 mM sodium phosphate at pH 8.0. After overnight incubation at 37°C, the reaction mixture was evaporated to 20 l using a SpeedVac and mixed with SDS sample buffer. The proteins were separated by SDS-PAGE, and the OmpB(AN) and OmpB(K) protein bands were excised from the gel and subjected to in-gel digestion followed by LC-MS/MS analysis as described (16).
Circular Dichroism and Fluorescence Measurements-Protein samples for CD and fluorescence measurements were prepared in 5 mM NaH 2 PO 4 and 50 mM KF at pH7.5. All CD measurements were performed between 180 and 280 nm at 25°C using a 1-mm cuvette and a Jasco J710 CD spectropolarimeter. The scanning speed used was 50 nm/min with an 8-s response time and 5-nm bandwidth. Each spectrum represents the average from five scans. The intrinsic fluorescence was measured using a FluoroMax-2 fluorometer at 25°C. Protein emission spectra were obtained with the excitation wavelength set at 280 nm using a cuvette with 10 mm-path length. Each spectrum obtained was the averaged of five scans between 300 and 450 nm with both the excitation and emission slits set at 5 nm.
Size Exclusion Chromatography with Multiangle Light Scattering-Wild type RpPKMT1 and RtPKMT2 were tested for their oligomeric state using size exclusion chromatography with multiangle light scattering (Wyatt Technology Corp.). Protein samples in 1ϫ PBS supplemented with tris(2-carboxyethyl)phosphine or DTT were separated by HPLC (Agilent 1200 series) using a size exclusion column (Shodex KW-803) equilibrated in 1ϫ PBS and with in-line UV, multiangle light scattering, and refractive index detectors (DAWN Heleos II and Optilab reX, respectively, Wyatt Technology Corp.) for molecular weight characterization. For each run, the protein sample in 100 l (2 mg/ml) was injected and eluted at a flow rate of 0.5 ml/min. UV and multiangle light scattering data were collected and analyzed using ASTRA software (Wyatt Technology Corp.). The calculated molecular weights of the peaks are presented as the mean with a 95% confidence interval (lower limit, upper limit). Bovine serum albumin (BSA; 66.5 kDa) was run as a control.