|
Observation of Covalent Intermediates in an Enzyme Mechanism at Atomic Resolution
Andreas Heine, Grace DeSantis, John G. Luz, Michael Mitchell, Chi-Huey Wong, and Ian A. Wilson
|
Supplementary Material
Active site residues in other class I aldolases
In FBP aldolases (
1), Lys229 on strand
6 forms the Schiff base, whereas in transaldolase B, Lys132 comes from strand
4 (Web Fig. 1). In FBP aldolases, Asp33 was originally hypothesized to protonate the incipient carbinolamine hydroxyl (C2) with Tyr363 abstracting the proton
to the carbonyl (C3 of DHAP). In less reactive isoenzymes, it was speculated that a water molecule might assume the latter role. However, in a recent study of human muscle aldolase (
2), Glu187 is now proposed as the catalytic acid and Asp33 as the catalytic base by analogy with the corresponding Glu96 and Asp17 of transaldolase B (
3). In
E. coli KDPG aldolase, Lys133 (
6) forms the Schiff base (
4) and Glu45 is proposed to act as a general base (
4,
5). Surprisingly, a K133Q/T161K double mutant still has some activity; Lys161 on strand
7 (Web Fig. 1) is now proposed to assume the Schiff base-forming role.
Substrate soaking
The catalytic mechanism for DERA has not previously been investigated. Our first attempts to co-crystallize the DRP substrate with the DERA enzyme did not lead to interpretable electron density maps for the ligand. The enzyme may have catalyzed the retroaldol reaction, resulting in a mixture of substrate/product in the crystal. We, therefore, decided to soak DRP into preformed crystals just prior to data collection and immediately flash-cool them. To address the question of appropriate diffusion time of ligand into the crystals, we performed a series of soaks lasting from 5 sec. to 10 min. (Web Fig. 2). With a soaking time of only 5 sec., electron density already started to appear for the heavier phosphate group. After 70 sec., the complete substrate molecule was visible in the electron density, with no further differences discernible in up to 10 min soaks. Strong connecting electron density between the Lys167 side chain and the substrate clearly identified a covalent intermediate in Schiff base formation. Furthermore, the substrate was found in its linear form, as in FBP aldolases (6, 7), and not as the cyclic furanose.
Superposition of aldolase enzymes
In order to look for conservation of catalytic residues in class I aldolase enzymes, we aligned the available coordinates (8) with the DERA wild-type carbinolamine complex using the superimposition matrix from DALI (9). Interestingly, with the exception of transaldolase, all the reactive lysines are not only on the same strand, but also in the same location (10) with respect to the center of the barrel (Web Fig. 1). The best overall alignment was obtained with KDPG aldolase (11), where their equivalent Lys167 (DERA) and Lys133 (KDPG) N
positions are only 2.3 Å apart. Similarly, the two substrate carbon atoms which form the covalent bond to their respective Lys-N are only 1.7 Å apart. The putative base Glu45 in KDPG overlaps with Cys47 (C
-S 0.5 Å apart), that is highly conserved in all DERA sequences. However, the KDPG Glu45 carboxyl Oe1 corresponds (1.4 Å apart) to a water molecule (Wat29) that could assume a similar catalytic role. For rabbit muscle aldolase (12), Lys229 overlaps closely with DERA Lys167 (1.2 Å N-N), but the adjacent Lys201 and Asp102 in DERA that could form a putative proton relay system are switched to Ser300 and Lys146, respectively. Cys47 (DERA) is now replaced by Ile77, whereas Ile139 (DERA) corresponds to the putative base Glu187 (13). Thus, a similar constellation of key catalytic residues appears to be present in these aldolase binding sites, but their precise configuration differs from enzyme to enzyme, suggesting that the aldolase mechanism does not require a unique geometric arrangement of catalytic residues.
Comparison of bond length and angles
Since we now have atomic resolution X-ray structures, we can compare the crystallographically-determined bond angles to their theoretically-expected values for the covalent intermediates in DERA and for the pyruvate adduct in KDPG aldolase (Web Table 2A). For the DERA wild-type carbinolamine complex, the observed angles agree well with the theoretical values and clearly indicate a tetrahedral arrangement around the C1 position of the substrate. These observed bond lengths and angles do not allow discrimination between a protonated, positively-charged e-amino group versus an uncharged neutral species, since the theoretical values of model compounds (InsightII, Molecular Structure, Inc) are almost identical. In the K201L mutant complex, a shorter N-C bond and an N-C-C angle of 125.0° for the Schiff base indicates a trigonal geometry at the carbon atom, in excellent agreement with the theoretical values. In contrast, the KDPG pyruvate-carbinolamine complex has a substantially more skewed geometry (5); the N-C-C angle of 79.4° is much smaller than expected, whereas the N-C-O angle with 140.3° is much larger. Furthermore, the limited resolution (1.95 Å) of the KDPG adduct and resulting electron density did not allow discrimination between the oxygen and carbon atoms (5); the identification of which atoms were O3 and C3 was based on hydrogen bonding interactions of O3 with Glu45 and Wat35. This covalent intermediate probably represents a non-productive adduct. Formation of the other enantiomer complex, where O3 and C3 are exchanged, would also seem feasible (14).
References and Notes
1. J. A. Littlechild, H. C. Watson, TIBS 18, 36 (1993).
2. A. Dalby, Z. Dauter, J. A. Littlechild, Protein Sci. 8, 291 (1999).
3. J. Jia, U. Schoerken, Y. Lindqvist, G. A. Sprenger, G. Schneider, Protein Sci. 6, 119 (1997).
4. N. Wymer et al., Structure 9, 1 (2001).
5. J. Allard, P. Grochulski, J. Sygusch, Proc. Nat. Acad. Sci. U.S.A. 98, 3679 (2001).
6. K. H. Choi et al., Biochemistry 38, 12655 (1999).
7. C. L. Verlinde, P. M. Quigley, J. Mol. Model 5, 37 (1999).
8. KDPG aldolase in complex with pyruvate (PDB code 1EUA) (5), human muscle fructose 1,6-bisphosphate aldolase (PDB code 4ALD) (2), rabbit muscle 1,6-bisphosphate fructose aldolase (PDB code 1ADO) (18) and transaldolase B (PDB code 1UCW) (3).
9. L. Holm, C. Sander, J. Mol. Biol. 233, 123 (1993).
10. For a more detailed pictorial comparison of reactive lysine positions in class I aldolase enzymes, equivalent sections of their -barrels (Web Table 1) were superimposed within InsightII (Molecular Simulations, Inc.). The superposition shows eight aldolase structure active sites.
11. For this alignment, the DALI Z-score is 12.4 with an rmsd of 2.7 Å based on 168 equivalent C positions. Fifteen fragments were used for the alignment and, for those residues, the sequence identity is 17%.
12. The DALI Z-score is 7.9; the rmsd based on 196 C positions is 3.5 Å.
13. The Z-score for human muscle aldolase is 12.2 with an rmsd of 3.2 Å based on 202 equivalent C positions. Again, Lys167 (DERA) and Lys229 superimpose very well and the two phosphor atom positions are 2.7 Å apart.
14. With only minor rearrangement, O3 would then be able to interact with Arg49 and Wat568 to form hydrogen bonds. C3 would then be in an ideal location for hydrogen removal, as it is then proximal to Glu45 and rotation around the lysine N- C bond as speculated (5) would not be required (and anyhow would result in a different location of the substrate carboxyl group from that found in the bound substrate).
15. Molecular Simulations Inc. Insight 2000. Setup as described in (19). The CVFF force field provided in the Discover module of InsightII was used.
16. P. J. Kraulis, J. Appl. Crystallog. 24, 946 (1991).
17. E. A. Merritt, D. J. Bacon, Meth. Enzymol. 277, 505 (1997).
18. N. Blom, J. Sygusch, Nat. Struct. Biol. 4, 36 (1997).
19. G. DeSantis, J.B. Jones, J. Am Chem.Soc. 120, 8582 (1998).
| Supplemental Table 1. Superposition of -barrels.
|
| Enzyme PDB | 1
6 res. | 2
6 res. | 3
7 res. | 4
6 res. | 5
6 res. | 6
7 res. | 7
6 res. | 8
6 res. | Reactive
lysine |
| DERA | 14-20 | 44-50 | 69-76 | 99-105 | 134-140 | 164-171 | 199-205 | 233-239 | 167 |
| 1eua | A16-A22 | A42-A48 | A68-A75 | A89-A95 | A109-A115 | A130-A137 | A157-A163 | A179-A185 | A133 |
| 1eun | A16-A22 | A42-A48 | A68-A75 | A89-A95 | A109-A115 | A130-A137 | A157-A163 | A179-A185 | A133 |
| 1fq0 | A16-A22 | A42-A48 | A68-A75 | A89-A95 | A109-A115 | A130-A137 | A157-A163 | A179-A185 | A133 |
| 1fwr | A16-A22 | A42-A48 | A68-A75 | A89-A95 | A109-A115 | A130-A137 | A157-A163 | A179-A185 | A161 |
| 1ado | A27-A33 | A74-A80 | A103-A110 | A143-A149 | A182-A188 | A226-A233 | A266-A272 | A297-A303 | A229 |
| 4ald | 27-33 | 74-80 | 103-110 | 143-149 | 182-188 | 226-233 | 266-272 | 297-303 | 229 |
| 1ucw | A14-A20 | A30-A36 | A91-A98 | A129-A135 | A151-A157 | A173-A180 | A221-A227 | A240-A246 | A132 |
Listed are the eight -barrels and the number of residues (res.) in each barrel used for the superposition.
B:
|
|
| Residue Atom | Structure, DRP
Carbinolamine | Minimized DRP
Carbinolamine | Minimized
Acetaldehyde Enamine |
| Asp102 Od2 | 5.33 Å | 5.42 Å | 5.26 Å |
| Lys201 N | 5.32 Å | 5.43 Å | 5.27 Å |
| Cys47 | 5.51 Å | 5.48 Å | 5.11 Å |
| Wat29 | 3.95 Å | 3.96 Å | 3.26 Å |
*Theoretical values are obtained from models built in InsightII and minimized using Discover (Molecular Simulations, Inc.).
Supplemental Figure 1. Comparison of class I aldolase active sites. A total of eight aldolase enzyme -barrel cores were superimposed in InsightII and their respective reactive lysine residues are shown. The DERA Lys167 is depicted in yellow (PDB code 1jcl), the three KDPG aldolase structures (PDB codes 1fq0, 1eua and 1eun) have their reactive lysine Lys133 represented in cyan, and the two FBP aldolases (PDB codes 1ado and 4ald) have their reactive lysine residues Lys229 represented in red. The only natural aldolase enzyme with its reactive lysine (green) on the fourth strand is transaldolase B (PDB code 1ucw). Even although the base of this lysine is shifted from the 6 to 4 strand, its N is in close proximity to all other aldolase lysine N's. The only outlier is Lys161, shown in blue for the double mutant K133Q/T161K of KDPG aldolase (4) (PDB code 1fwr). Here, the reactive lysine is located on 7 and, in addition, points away from the center of the barrel. However, the accurate position of the terminal atoms of Lys161 is uncertain, since the final electron density appears to be weak (4) and the PDB coordinates have zero occupancies for Ce and N for Lys161. It could be speculated that the weak activity for this double mutant could be partly due to a background reaction occurring within the active site.
Medium version | Full size version
Supplemental Figure 2. Substrate diffusion into the crystals. Four DERA-substrate complexes were determined at different DRP substrate soaking times of 5, 15, 70 sec. and 10 min. Electron density for the heavier phosphate group already appears after 5 sec. Continuous density is observed between Lys167 and the substrate molecule, indicating formation of a covalent intermediate. From the 3Fo-2Fc electron density contoured at 2 
(blue) and 4 (purple), it is apparent that the substrate is completely bound after a 70 sec. soak with almost no difference compared to a soak of 10 min. The electron density for Lys167 is displayed for comparison with substrate density. All structural figures were made with Bobscript (16) and Raster3D (17).
Medium version | Full size version
