A New UAG-Encoded Residue in the Structure of a Methanogen Methyltransferase

See allHide authors and affiliations

Science  24 May 2002:
Vol. 296, Issue 5572, pp. 1462-1466
DOI: 10.1126/science.1069556


Genes encoding methanogenic methylamine methyltransferases all contain an in-frame amber (UAG) codon that is read through during translation. We have identified the UAG-encoded residue in a 1.55 angstrom resolution structure of the Methanosarcina barkerimonomethylamine methyltransferase (MtmB). This structure reveals a homohexamer comprised of individual subunits with a TIM barrel fold. The electron density for the UAG-encoded residue is distinct from any of the 21 natural amino acids. Instead it appears consistent with a lysine in amide-linkage to (4R,5R)-4-substituted-pyrroline-5-carboxylate. We suggest that this amino acid be named l-pyrrolysine.

The catabolism of methylamines by methanogens involves a conserved arrangement of proteins. A specific monomethylamine (MMA), dimethylamine (DMA), or trimethylamine (TMA) methyltransferase activates the substrate for methyl transfer to a cognate corrinoid protein (1, 2). A second methyltransferase catalyzes the transfer of the methyl group from the methylated corrinoid cofactor to coenzyme M (CoM). Methyl-CoM is subsequently used to generate methane by methyl-CoM reductase (3).

All known methanogen methylamine (MMA, DMA, or TMA) methyltransferase genes contain a single in-frame amber (UAG) codon that does not appear to stop translation during protein synthesis (4,5). This strict conservation contrasts with the lack of sequence similarity between different MMA, DMA, and TMA methyltransferase gene families. Analysis of tryptic fragments of MMA methyltransferase (MtmB) by mass spectrometry and Edman degradation suggested that the amber codon serves as a sense codon that corresponds to lysine (6). The harsh conditions of peptide isolation, however, left open the possibility that the amber codon signals a modified lysine residue. The use of what is normally a stop codon to signal the incorporation of an unusual amino acid has precedent in the use of UGA to encode selenocysteine, the 21st natural amino acid found in Bacterial, Eucaryal, and Archaeal proteins (7, 8).

The structure of the Methanosarcina barkeri MS monomethylamine methyltransferase was determined to clarify the identity of the UAG-encoded residue (Tables 1 and 2). Two forms of the enzyme were obtained from crystallization conditions that differed only in the precipitating salt used [NaCl for form 1, and (NH4)2SO4 for form 2] and were solved to 1.55 Å and 1.7 Å resolution, respectively. The global conformations for these two forms are virtually identical, except for the region around the UAG-encoded amino acid. The overall MtmB structure consists of a homohexamer arranged into a dimer of trimers with overall D3 symmetry (Fig. 1A). Each subunit adopts an α/β TIM barrel fold (9, 10) (Fig. 1B) that is reminiscent of other corrinoid-cofactor associated proteins: tetrahydrofolate:corrinoid/iron-sulfur protein methyltransferase (11), diol dehydratase (12), and racemases/mutases (13, 14), although no significant sequence similarity is found between MtmB and these proteins. The eight-stranded β barrel is formed from strands β6 to β13 (15). Consistent with other methyltransferases, this barrel forms a deep cavity that for the MtmB is negatively charged (15). This may facilitate binding of the methylammonium cation. The location of the amber-encoded residue at the bottom of this cavity suggests its possible role in catalysis.

Figure 1

(A) Ribbon diagram of the MtmB hexamer. The subunits forming the two trimers are shaded in red and blue hues, respectively. (B) Ribbon diagram of one MtmB subunit colored by secondary structure element (α helices, red; β sheets, cyan; random coil, green). The atoms of the UAG-encoded residue are shown as ball-and-stick models and are colored by their elements, with carbon as gray, nitrogen as blue, and oxygen as red. These figures were prepared with the programs MOLSCRIPT, and Raster3D (22,23).

Table 1

Statistics for data collection, and phase determination. The MtmB crystals contain one MtmB subunit per asymmetric unit and belong to space group P6322 with unit cell dimensions a = b = 158.8 Å,c = 136.5 Å. The details regarding the crystallization, data collection, and structure determination are provided as supplementary material (15). All the calculations were done with the program PHASES (20). For completeness and R sym, numbers in parentheses represent the statistics for the shell comprising the outer 10% (theoretical) of the data. Phasing power is the mean value of the heavy atom structure factor divided by the lack of closure. Figure of merit is the mean value of the cosine of the error in phase angles. The combined figure of merit for all datasets was 0.63.

View this table:
Table 2

Refinement statistics. The starting R and Free-R factors after the first rigid body refinement were 0.45 and 0.44. The model was refined by simulated annealing, followed by several cycles of minimization using a maximum likelihood target based on the amplitudes and incorporation of an flat bulk solvent correction and an overall anisotropic B factor. The quality of the final model of the MtmB was assessed using the program PROCHECK and was found to be acceptable. The statistics for each of the stereochemical parameters was inside or better than the expected values, and 90% of the residues were found to occupy the most favored regions of the Ramachandran plot (21). The refined model consists of all the residues from 2 to 458, except the NH2-terminal methionine. This is consistent with its posttranslational cleavage as determined by NH2-terminal sequencing of the isolated protein (4). There are a total of three residues having thecis conformation, Pro204, Glu235, and Val367. rms, root mean square.

View this table:

The structure of MtmB from both NaCl and (NH4)2SO4 crystal forms suggests that the amber-encoded residue is distinct from the 21 known natural amino acids. Both structures support its biochemical assignment as a lysine core modified by a group attached to its epsilon nitrogen (Fig. 2). The identity of the modifying group appears to be (4R,5R)-4-substituted-pyrroline-5-carboxylate, with the carboxylate of the modifying group in amide linkage to the epsilon nitrogen of lysine (Fig. 2B). In both crystal forms, the modifying group is disordered between two distinct orientations of the pyrroline ring (15). Although this disorder is present in both crystal forms, the relative occupancy of each orientation in each crystal form differs.

Figure 2

(A) Fit of (4R,5R)-4-substituted-pyrroline-5-carboxylate to the 2F OF C 3σ density of the NaCl crystal form (orientation 1). The substituent attached to the C-4 carbon is shown as a methyl, but it could also be an ammonium or hydroxyl group. (B) Stick-diagram of proposedl-pyrrolysine amino acid. (C) ResidualF OF C difference map of NH4SO4 crystal form after incorporation of a 40% occupancy model consisting of l-pyrrolysine in orientation 1 and an exogenous ammonium ion. This remaining omit density suggests that l-pyrrolysine adopts a different orientation (orientation 2) at 60% occupancy in NH4SO4 with an amine added to the C-2 carbon of the pyrroline ring. These figures were prepared with the programs XtalView, MOLSCRIPT, and Raster3D (22–24).

The initial assignment of the UAG-encoded amino acid was based on the 1.55 Å resolution structure of the NaCl form. This form gave the clearest density because one orientation of the ring dominates (orientation 1). The initial 5σ F OF C density could be fit to a residue resembling β-methyl-D-proline in amide linkage with the epsilon nitrogen of lysine (Figs. 2A and 3A). Position 1 in the five-membered ring was assigned to a nitrogen atom because it forms hydrogen bonds with the carboxylate side chains of Glu259and Glu229, and because its refinement as nitrogen gave a better fit to the density than either carbon or oxygen. The identity of the atom attached to the C-4 ring carbon is currently unclear. Based on the current fit to the electron density, it could be a methyl, ammonium, or hydroxyl group. It is within hydrogen-bonding distance of Tyr335, but this distance is long, 3.16 Å, and two waters form a hydrogen bond to Tyr335 with a better geometry.

Figure 3

Stereoview of primary forms of the active site around the amber-encoded ligand: (A) NaCl crystals; (B) (NH4)2SO4 crystals, 40% occupancy orientation that is similar to NaCl crystals; (C) (NH4)2SO4 crystals, 60% occupancy orientation with amine added to ring. These figures were prepared with the programs XtalView, MOLSCRIPT, and Raster3D. (22–24).

After refinement of the initial model built from the 5σ electron density, weak but broad 3σ difference density was observed (15). This residual density suggested that the five-membered ring also adopts a second orientation (orientation 2), though at much lower occupancy. In this second orientation, this ring is rotated by approximately 90° relative to its position in orientation 1 and has different set of hydrogen bonding interactions with the protein. Refinement of the occupancies and thermal parameters of the modifying group in both orientations suggests that the relative occupancies of orientations 1 and 2 in the NaCl crystal form are 85% and 15%, respectively.

The 2F OF C density for UAG-encoded residue of the (NH4)2SO4 crystal form (1.7 Å resolution) differs significantly from that observed in the NaCl crystal form (15). It can be initially fit, however, to a model composed of the same two orientations found in the NaCl crystal form, although with different relative occupancies (15). Refinement suggests that in the (NH4)2SO4 crystal form, orientation 2 now becomes the dominant conformation, with the relative occupancies of orientations 1 and 2 having values of 40% and 60%, respectively.

In addition to the difference in the relative occupancies, electron density maps of the (NH4)2SO4crystal form also reveal the presence of an atom bound to the C-2 ring carbon for orientation 2 that was not observed in the NaCl crystal form (Figs. 2C and 3C). We assign this atom to nitrogen, because (NH4)2SO4 is the only salt not present in the solution used to grow the NaCl crystals. Consistent with this assignment, the side chains of Glu259 and Gln333 are at hydrogen-bonding distances to the proposed amine substituent. These interactions likely stabilize thel-pyrrolysine in orientation 2, thereby accounting for its higher occupancy in the (NH4)2SO4crystal form.

In order to account for some remaining difference density in the (NH4)2SO4 crystal form, a free ammonium ion was required at 40% occupancy (15). This likely represents enzyme-bound ammonium ion that is associated with the protein when the ring is in orientation 1 (Fig. 3B). This ion does not bind or interact with the UAG-encoded residue, but instead forms a hydrogen bond with Met261 and Tyr335. The ammonium ion occupies a position near the amine that is added to the C-2 carbon of the ring in orientation 2 of the (NH4)2SO4 crystal form. This structure may reflect an intermediate state before amine addition.

In order for an amine to add to the ring, the C-2 carbon must be sp2 hybridized. This hybridization is likely achieved by having an imine bond between the N-1 and C-2 atoms of the ring. This implies that the identity of the UAG-encoded amino acid before amine addition is (4R,5R)-4-substituted-pyrroline-5-carboxylate in amide linkage to the epsilon nitrogen of lysine, a species that we proposed be named l-pyrrolysine (Fig. 2B). We note that although the modifying group has analogy to Δ1-pyrroline-5-carboxylate, the direct precursor for proline synthesis, the D chirality of the pyrroline-5-carboxylate and the additional substituent at the C4 carbon make it distinct.

MtmB in complex with its associated corrinoid protein, MtmC, was subjected to electrospray mass spectrometry to confirm the pyrrolysine assignment (16). The measured masses of MtmB and MtmC were 50,105 ± 2 daltons and 23,066 ± 1 daltons, respectively. These values can be compared with the deduced mass from the predicted protein sequences from the encoding genes (4) using the program SHERPA (17). Without the corrinoid prosthetic group, the theoretical average mass of MtmC is calculated as 23,067, which is consistent with the experimentally derived molecular mass. For MtmB, the theoretical average mass is calculated as 49,998, by assuming a lysine residue at the UAG-encoded position. This calculated value is 107 atomic mass units (amu) less than the experimental mass of MtmB (50,105 daltons). With the incorporation of the 4-substituted-pyrroline-5-carboxylate group as found in the crystal structure, however, the theoretical molecular mass for MtmB increases by 109 amu to 50,107 daltons assuming a methyl group is attached to the C-4 ring carbon. If the substituent attached to the ring is either an amine or hydroxyl group, then the theoretical mass would be higher by 110 amu (50,108 daltons) and 111 amu (50,109 daltons), respectively.

MtmB assists in the transfer of the methyl group of monomethylamine to the corrinoid cofactor of MtmC. A mechanism for how pyrrolysine could play a role in activation of methylamine substrates is suggested on the basis of the two forms of the enzyme (Fig. 4). In this model, the role of pyrrolysine is to position and display the methyl group of methylamine for attack by the corrinoid cofactor. A similar mechanism could also be envisioned for the TMA and DMA methyltransferases, which also have UAG-encoded residues.

Figure 4

Hypothetical model for the role of the amber-encoded residue in catalysis. The proposed intermediates for (A), (B), (C), and (F) are based on the structures ofl-pyrrolysine (X = Me, NH2, or OH) in its 85% occupancy orientation in NaCl crystal form and on both orientations of l-pyrrolysine in the (NH4)2SO4 crystal form. Intermediates (D) and (E) are based on a preliminary docking model of MtmB with its cognate corrinoid protein, MtmC.

The two conformations of pyrrolysine and the surrounding side chains are consistent with induced fit mechanisms observed in other systems. Here, the two orientations of the ring may facilitate different interactions with the carboxylate side chain of Glu259. We propose a mechanism in which the carboxylate of Glu259 is protonated before methylamine addition and positioned so that the carboxyl hydrogen bonds to the pyrroline ring nitrogen of pyrrolysine. Protonation of the imine nitrogen would make it more electron withdrawing and help to activate the C-2 carbon for nucleophilic addition. After methylamine addition to the C-2 carbon, the new deprotonated carboxylate of Glu259 shifts to be a hydrogen bond acceptor with the bound methyl ammonium group. The amide side chain of Gln333 provides a second hydrogen bond acceptor to the bound methyl ammonium group. These two hydrogen bonds and the covalent bond to the pyrroline ring serve to position the methylammonium group so that its methyl group is directed toward the surface of the binding cleft. Here, it presumably is positioned to interact with the corrinoid-cofactor of MtmC upon formation of the MtmBC complex. Preliminary docking models of MtmB with the B12-binding domain of MetH (18), a homolog of MtmC (4) have indicated the feasibility of this concept. Additional work will be required to confirm this mechanism and our proposed structure for l-pyrrolysine; however, these models provide fertile grounds for future experimentation.

In conclusion, the in-frame amber codon in the gene encoding MtmB encodes a novel amino acid with properties favoring catalysis of methyltransferase reactions involving amines. As described in the accompanying paper (19), an unusual aminoacyl-tRNA synthetase and tRNACUA are encoded near the mtmBgenes in M. barkeri. Although it formally remains possible that modification of lysine occurs after UAG directed insertion of lysine into the protein, the use of a canonical stop codon and dedicated tRNA is most consistent with the direct translational encoding of l-pyrrolysine. Thus current evidence supports the case for l-pyrrolysine representing the 22nd naturally occurring amino acid to be identified.

  • * To whom correspondence should be addressed: E-mail: chan{at} and krzycki.1{at}


View Abstract

Navigate This Article