The catalytic cycle of β-lactam synthetase observed by x-ray crystallographic snapshots

Abstract

The catalytic cycle of the ATP/Mg²⁺-dependent enzyme β-lactam synthetase (β-LS) from Streptomyces clavuligerus has been observed through a series of x-ray crystallographic snapshots. Chemistry is initiated by the ordered binding of ATP/Mg²⁺ and N²-(carboxyethyl)-l-arginine (CEA) to the apoenzyme. The apo and ATP/Mg²⁺ structures described here, along with the previously described CEA⋅α,β-methyleneadenosine 5′-triphosphate (CEA⋅AMP-CPP)/Mg²⁺ structure, illuminate changes in active site geometry that favor adenylation. In addition, an acyladenylate intermediate has been trapped. The substrate analog N²-(carboxymethyl)-l-arginine (CMA) was adenylated by ATP in the crystal and represents a close structural analog of the previously proposed CEA-adenylate intermediate. Finally, the structure of the ternary product complex deoxyguanidinoproclavaminic acid (DGPC)⋅AMP/PP_i/Mg²⁺ has been determined. The CMA-AMP/PP_i/Mg²⁺ and DGPC⋅AMP/PP_i/Mg²⁺ structures reveal interactions in the active site that facilitate β-lactam formation. All of the ATP-bound structures differ from the previously described CEA⋅AMP-CPP/Mg²⁺ structure in that two Mg²⁺ ions are found in the active sites. These Mg²⁺ ions play critical roles in both the adenylation and β-lactamization reactions.

The characterization of natural product biosynthetic pathways at the genetic level has revealed a growing number of examples in which the synthetic capability of a class of enzymes has been directed to a different end (1, 2). A particularly striking instance is the relationship between Streptomyces clavuligerus β-lactam synthetase (β-LS) and the universally distributed and highly conserved family of asparagine synthetases, class B (AS-Bs). Despite an overall sequence homology, the two enzymes catalyze quite different reactions. β-LS forms the four-membered ring of the potent β-lactamase inhibitor clavulanic acid (Fig. 1; refs. 3 and 4). Clavulanic acid is of increasing clinical importance for its ability to overcome bacterial resistance to penicillins and cephalosporins (5, 6) by inactivating β-lactamase enzymes (7). AS-Bs convert aspartic acid to asparagine (ref. 8; Fig. 1).

Fig 1.

Reactions catalyzed by β-LS and AS-B. β-LS cyclizes CEA to DGPC. Reactions catalyzed by other enzymes result in the formation of clavulanic acid. The truncated substrate analog CMA is a dead-end inhibitor. AS-B converts glutamine and aspartic acid to glutamic acid and asparagine.

Recent crystal structures of both enzymes reveal the evolutionary changes that differentiate β-lactam formation from asparagine synthesis. AS-B from Escherichia coli comprises two catalytic domains (9). The smaller N-terminal domain houses a glutaminase site with a catalytically essential cysteine residue. Glutamine is hydrolyzed to glutamic acid and ammonia, which is believed to traverse a tunnel extending through the domain interface into the larger C-terminal domain. A second catalytic center residing in this domain mediates the adenylation of aspartic acid and its subsequent reaction with ammonia to form asparagine. By contrast, β-LS, which resembles AS-B in overall structure (10), has a less organized N-terminal domain and lacks the strictly conserved cysteine critical to glutaminase function (3). Like AS-B, β-LS adenylates its substrate, N²-(carboxyethyl)-l-arginine (CEA; Fig. 1), but catalyzes intramolecular acyl substitution by nitrogen rather than intermolecular reaction with ammonia (Fig. 1). The C-terminal active site is significantly larger in β-LS, reflecting the size difference between CEA and aspartic acid (10).

β-LS is an excellent candidate for protein engineering experiments aimed at constructing new β-lactamase inhibitors and β-lactam antibiotics. Based on kinetic data, the chemical mechanism of β-LS has been proposed to involve adenylation of CEA followed by intramolecular β-lactamization via a tetrahedral intermediate (11). Consistent with this mechanism, the crystal structure of β-LS determined in the presence of the substrate CEA, the ATP analog α,β-methyleneadenosine 5′-triphosphate (AMP-CPP), and a Mg²⁺ ion reveals that the unreacted substrates are preorganized in the active site to favor these reactions (10). To fully exploit the synthetic potential of β-LS, we have determined its structure in the absence of any substrates or cofactors (apo), in complex with ATP/Mg²⁺, in complex with ATP/Mg²⁺ and the dead-end inhibitor N²-(carboxymethyl)-l-arginine (CMA; Fig. 1; ref. 11), and in complex with the products deoxyguanidinoproclavaminic acid (DGPC), AMP, and PP_i/Mg²⁺. These four structures provide details about protein–substrate interactions that open for consideration alternative mechanisms of β-lactam formation.

Materials and Methods

Crystallization and Data Collection.

Purification and crystallization of S. clavuligerus β-LS was carried out as described (10) by using precipitant solutions containing 7–10% polyethylene glycol 4000, 200 mM MgCl₂, 15% glycerol, and 80 mM Tris (pH 8.0). The final MgCl₂ concentration in the drops was 100 mM. Soaking experiments were performed in a solution containing 24% (wt/vol) polyethylene glycol 4000, 260 mM MgCl₂, 20% (vol/vol) glycerol, and 80 mM Tris (pH 8.0). Crystallization, soaking, and all buffer preparation were carried out at room temperature. The soaking solution was supplemented with three different combinations of ligands: 5 mM ATP, 5 mM CMA plus 5 mM ATP, and 5 mM DGPC plus 5 mM AMP plus 5 mM PP_i. After soaking in these solutions for 1–2 h, crystals were flash cooled in liquid nitrogen. ATP, AMP, and PP_i were obtained from Sigma, and CMA (11) and DGPC (3) were obtained as described. All of the soaked crystals belonged to the space group P2₁, and there are two β-LS monomers (labeled A and B) in the asymmetric unit. According to gel filtration and native gel analyses, β-LS is a dimer in solution at high concentrations (10). Data were processed with the programs denzo and scalepack (Table 1; ref. 12).

Table 1.

Data collection and refinement statistics

	apo	ATP/Mg2+	CMA-AMP/PPi/Mg2+	DGPC⋅AMP/PPi/Mg2+
Data collection
Unit cell dimensions
a, Å	61.2	61.3	61.4	61.0
b, Å	97.5	97.6	97.2	96.6
c, Å	81.0	81.3	81.1	80.9
β, °	90.4	90.7	90.1	90.5
Resolution range, Å	24–1.95	25–2.10	29–2.47	29–2.16
No. of total observations	690,463	594,199	382,028	687,896
No. of unique observations	69,211	54,280	34,018	50,318
Completeness (%)	92.8 (90.0)	99.6 (98.5)	99.1 (99.1)	97.4 (93.1)
Rsym	0.051 (0.253)	0.084 (0.296)	0.094 (0.292)	0.069 (0.293)
% > 3σ (I)	83.1 (55.3)	74.3 (44.7)	73.6 (41.4)	66.1 (30.0)
Refinement
Resolution range, Å	24–1.95	25–2.11	29–2.47	29–2.16
No. of reflections	63,190	53,056	32,528	47,767
R-factor‡	19.7	21.0	21.4	20.8
Rfree§	23.1	25.1	27.0	24.7
No. of atoms
Protein, nonhydrogen	7,282	7,399	7,449	7,446
Nonprotein	603	568	475	505
rms bond length, Å	0.007	0.013	0.028	0.010
rms bond angles, °	1.5	2.0	1.8	1.7
Average B value, Å2	28.4	22.7	20.0	26.4

 Data were collected at −160°C by using a Mar 345 imaging plate detector at beamline 9-1 at Stanford Synchrotron Radiation Laboratory (SSRL). Wavelength, 1.000 Å.

^*Values in parentheses are for the highest resolution shells.

^† R_sym = Σ|I_obs − I_avg|/ΣI_obs, where the summation is over all reflections.

^‡ R-factor = Σ|F_obs − F_calc|/ΣF_obs.

^§Test set size: apo, 9.0%; ATP/Mg²⁺, 7.8%; CMA-AMP/PP_i/Mg²⁺, 9.3%; DGPC⋅AMP/PP_i/Mg²⁺, 7.8%.

Structure Determination.

The structure of β-LS complexed with the substrate CEA and the ATP analog AMP-CPP (CEA⋅AMP-CPP/Mg²⁺, PDB ID code ; ref. 10) was used as a starting model for refinement of each structure. In addition to the three soaking experiments described above, the structure of β-LS in the absence of any exogenous ligands (apo) was determined (Table 1). The structures were refined with the program CNS (13) by rigid body refinement followed by iterative cycles of simulated annealing and individual B-value refinement and model adjustment with the program xtalview (14). The resolutions of the structures range from 1.95 Å for the apo structure to 2.47 Å for the CMA-AMP/PP_i/Mg²⁺ structure (Table 1). The same test set, originally generated for the apo data set, was used for cross validation of each structure. The final model for the apo structure consists of residues A3–A20, A25–A164, A170–A443, A454–A467, A473–A506, B2–20, B24–B358, B363–B441, B452–B508, and 603 water molecules. For the ATP/Mg²⁺ structure, residues A4–A19, A23–A163, A170–A443, A454–A507, B3–B20, B24–B445, B452–B508, 470 water molecules, 4 Mg²⁺ ions, 1 ATP molecule at full occupancy, and 1 ATP molecule, 1 AMP molecule, 1 PP_i molecule, all at half occupancy, were modeled. The final model for the CMA-AMP/PP_i/Mg²⁺ structure contains residues A3–A19, A25–A164, A169–A507, B3–B20, B25–B444, B453–B508, 365 water molecules, 4 Mg²⁺ ions, 2 CMA-AMP adenylated intermediate molecules, 2 PP_i molecules, and 2 glycerol molecules. For the DGPC⋅AMP/PP_i/Mg²⁺ structure, residues A5–A17, A25–A163, A169–A507, B3–B444, B453–B508, 405 water molecules, 4 Mg²⁺ ions, 2 AMP molecules, 2 PP_i molecules, and 2 DGPC molecules were modeled. No electron density was observed for the amino acid residues not modeled in each structure. Ligands were modeled into difference electron density observed in the active sites, and appropriate refinement restraints were calculated by using the HIC-Up server (15). Ramachandran plots calculated with procheck (16) indicate that the four models exhibit good geometry with all residues in the most favored or additionally allowed regions. Figures were generated with molscript (17), raster3d (18), and bobscript (19).

Results

ATP/Mg²⁺ Structure.

The active site in monomer B of the ATP soaked crystal contains a well defined ATP molecule (Figs. 2A and 3A). The ATP occupies the same position as AMP-CPP in the original β-LS CEA⋅AMP-CPP/Mg²⁺ structure (ref. 10; Fig. 3B) with its adenosine hydroxyl groups hydrogen bonded to the amide nitrogens of Gly-347 and Tyr-348 and the carbonyl oxygen atom of Val-247 and its adenine nitrogens hydrogen bonded to the carbonyl oxygen atom and amide nitrogen of Met-273. There are important differences between the structures, however. In the CEA⋅AMP-CPP/Mg²⁺ structure, one Mg²⁺ ion is present whereas in the ATP/Mg²⁺ structure, two Mg²⁺ ions are observed. The first, labeled Mg1, corresponds to the Mg²⁺ ion in the CEA⋅AMP-CPP/Mg²⁺ structure, but the coordination is different. As previously observed, it is coordinated by one side chain oxygen atom each from Asp-253 and Asp-351, a water molecule, and terminal oxygen atoms from the α- and γ-phosphates. Instead of the terminal β-phosphate oxygen atom, however, the sixth ligand is the oxygen atom bridging the α- and β-phosphates. The second Mg²⁺ ion, labeled Mg2, is located on the opposite side of the ATP phosphorus atoms and coordinated by terminal oxygen atoms from the α-, β-, and γ-phosphates and by three water molecules. Two of the coordinated water molecules are hydrogen bonded to the side chain oxygen atoms of Glu-280, and the third interacts with the carbonyl oxygen atom of Lys-443 via an intervening water molecule. The coordination geometry at both sites is approximately octahedral. Besides coordinating the Mg²⁺ ions, the β-phosphate oxygen atoms interact with Ser-249 and Ser-254, and the γ-phosphate oxygen atoms interact with Lys-423 and Lys-443. The absence of the second Mg²⁺ ion in the CEA⋅AMP-CPP/Mg²⁺ structure is likely due to the substitution of a carbon atom for the oxygen atom bridging the α- and β-phosphates. Two Mg²⁺ ions are also observed in monomer A, but the electron density is less well defined and was best modeled as a mixture of ATP and AMP plus PP_i, both at half occupancy.

Fig 2.

Electron density maps showing bound ligands in the three soaked crystals. The Mg²⁺ ions are shown as gray spheres. (A) ATP/Mg²⁺ structure. Stereo view of a simulated annealing omit map contoured at 2σ with the ATP and Mg²⁺ ions omitted. (B) CMA-AMP/PP_i/Mg²⁺ structure. Stereo view of a simulated annealing omit map contoured at 2σ with the CMA-adenylate, PP_i, and Mg²⁺ ions omitted. (C) DGPC⋅AMP/PP_i/Mg²⁺ structure. Stereo view of a simulated annealing omit map contoured at 2σ with the DGPC, AMP, and PP_i omitted.

Fig 3.

Stereo views of ligands bound in the β-LS active site. Only amino acid residues directly coordinated to the Mg²⁺ ions are shown. Solvent molecules are shown as red spheres. (A) ATP/Mg²⁺ structure. (B) Previously determined CEA⋅AMP-CPP/Mg²⁺ structure. (C) CMA-AMP/PP_i/Mg²⁺ structure. (D) DGPC⋅AMP/PP_i/Mg²⁺ structure.

CMA-AMP/PP_i/Mg²⁺ Structure.

The substrate analog CMA is a competitive inhibitor of β-LS with respect to CEA (11). In both monomers in the asymmetric unit, electron density corresponding to CMA and ATP is present. Unexpectedly, the electron density clearly indicates that the CMA molecule has reacted with the ATP, producing CMA-adenylate and PP_i (Figs. 2B, 3C, and 4). This structure thus captures a key intermediate in the β-LS reaction cycle. The interactions between CMA and the protein are the same as those observed in the CEA⋅AMP-CPP/Mg²⁺ structure (10). The guanidino group is hydrogen bonded to Glu-382 and Asp-373, and the α-carboxylate interacts with the amide nitrogen of Gly-349 and a water molecule coordinated to Mg1. This similarity is consistent with the observation that the binding constants for CMA and CEA are the same order of magnitude (11). Both Mg²⁺ ions are still present, and although the ATP has been converted to CMA-AMP and PP_i, the Mg coordination is the same as in the ATP/Mg²⁺ structure, with both Mg²⁺ ions bound to the CMA-AMP phosphate and to each phosphate of PP_i. The third water molecule coordinated to Mg2 has been replaced by the carbonyl oxygen atom of Leu-444, however. In the apo, ATP/Mg²⁺, and previous CEA⋅AMP-CPP/Mg²⁺ structures, most of the loop encompassing residues 444–453 is disordered. This loop has become ordered in monomer A of the CMA-AMP/PP_i/Mg²⁺ structure, however, and blocks exposure of the CMA-adenylate to solvent.

Fig 4.

Mechanistically important hydrogen bonding interactions in the CMA-AMP/PP_i/Mg²⁺ active site. The conformations of Tyr-326, Tyr-348, and Lys-443 differ from those observed in the apo structure (superimposed in green).

The interactions between the phosphate oxygen atoms and Ser-249, Ser-254, Lys-423, and Lys-443 are the same as those observed in the ATP/Mg²⁺ structure. In addition, the α-phosphate oxygen atoms in the adenylated intermediate interact with Lys-443 and the amide nitrogen of Val-446, and one phosphate oxygen atom from PP_i interacts with the amide nitrogen of Leu-444. The β-carboxylate of the CMA-adenylate is ≈3.4 Å from the side chain of Lys-443 (Fig. 4). Finally, addition of CMA alters the conformation of two residues in the active site, Tyr-326 and Tyr-348. In the apo and ATP/Mg²⁺ structures, these tyrosines point away from the substrate binding site. On CMA binding, both rotate toward the bound substrate analog (Fig. 4). The change is most drastic for Tyr-326, which partially covers the CMA molecule, securing it in the active site cleft. These two residues occupy similar positions in the CEA⋅AMP-CPP/Mg²⁺ structure (10).

DGPC⋅AMP/PP_i/Mg²⁺ Structure.

It is also possible to soak all of the reaction products into the β-LS active site. Both active sites in the asymmetric unit contain DGPC, AMP, and PP_i (Figs. 2C and 3D). The active site is very similar to that observed in the presence of CMA-adenylate and PP_i. The two tyrosines remain shifted, partially covering the DGPC, and residues 444–453 are ordered in the A monomer. The guanidino and α-carboxylate groups of DGPC participate in the same hydrogen bonding interactions, and the coordination of the two Mg²⁺ ions, as well as interactions between the phosphate oxygen atoms and the protein, is unchanged after β-lactam formation. Finally, the carbonyl oxygen atom of the β-lactam ring is hydrogen bonded to the side chain of Lys-443.

Discussion

A two-stage reaction is catalyzed by β-LS, adenylation, and β-lactam formation (Fig. 5). Adenylation of the CEA β-carboxylate, which occurs by in-line displacement of the phosphoric anhydride of ATP by a carboxylate oxygen atom, is a simple substitution reaction, and no net charge is created during this step at physiological pH. Negative charge has simply migrated from the β-carboxylate to PP_i. β-lactam formation is more complex. Intramolecular nucleophilic addition of the α-amine of CEA to the activated β-carboxylethyladenylate would result in the formation of an oxyanion intermediate or transition state, which undergoes α-elimination, expelling AMP/Mg²⁺ to form the β-lactam ring (Fig. 5). For acyl substitution reactions involving very good leaving groups, as is the case here, concerted mechanisms are believed to occur through a tetrahedral transition state (20, 21), or even by a dissociative mechanism and formation of an acylium ion (22, 23). For nitrogen nucleophiles, however, the development of positive charge on nitrogen as bond formation to the carbonyl carbon proceeds (Fig. 5) gives a more ambiguous situation where existing data do not distinguish between formation of a tetrahedral transition state or intermediate (20, 23). Notwithstanding, the crystal structures show that β-LS has evolved to provide functional groups in the active site that both optimize reaction geometries, and stabilize intermediates and transition states during the course of the β-LS reaction sequence.

Fig 5.

Chemical mechanism of β-LS based on structural and kinetic data.

In the first step of the catalytic cycle, ATP and two Mg²⁺ ions bind to the ligand-free form of the enzyme (Fig. 3A). The substrate CEA binds next (Fig. 3B), and is clamped into place by Tyr-326 and Tyr-348, which adopt different conformations in the CEA⋅AMP-CPP/Mg²⁺, CMA-AMP/PP_i/Mg²⁺, and DGPC⋅AMP/PP_i/Mg²⁺ structures than in the apo and ATP/Mg²⁺ structures (Fig. 4). This order of substrate binding is supported by CMA inhibition studies (11). The CEA⋅AMP-CPP/Mg²⁺ structure shows a high degree of substrate preorganization, with the CEA β-carboxylate poised to attack the α-phosphate of AMP-CPP (Fig. 3B; ref. 10). The CEA-adenylate intermediate has not been directly observed in solution (11), but is modeled well by the CMA-adenylate observed in the CMA-AMP/PP_i/Mg²⁺ structure (Fig. 3C). This intermediate is captured in the crystal presumably because cyclization of CMA to a three-membered ring involves a high energy barrier. The adenylation step is facilitated by both Mg²⁺ ions. The shift of Mg1 from the terminal β-phosphate oxygen in the CEA⋅AMP-CPP/Mg²⁺ structure (Fig. 3B) to the bridging α,β-oxygen of the triphosphate in ATP (Fig. 3A) draws on the Lewis acidity of Mg²⁺ to further activate the α,β-bond for direct displacement by the CEA β-carboxylate, amplifying the effect of metal coordination on the adenylation reaction.

The presence of Mg2 in all of the structures except that of CEA⋅AMP-CPP/Mg²⁺ is striking, but not unprecedented. A number of enzymes, including the MutT pyrophosphohydrolase (24, 25), require two divalent cations. In addition, two Mg²⁺ ions are observed in the crystal structures of glutathione synthetase (26) and the MurD (27) and d-Ala-d-Ala (28) ligases. The two Mg²⁺ ions are of central importance to the β-LS catalytic mechanism. Together, they balance the overall charge in the active site and stabilize negative phosphate charges, in particular, the developing negative charge during adenylation. Further stabilization is provided by interaction of the PP_i anion with Lys-423 and Lys-443.

The next step is formation of the four-membered β-lactam ring. The CMA-AMP/PP_i/Mg²⁺ and DGPC⋅AMP/PP_i/Mg²⁺ structures represent time points bracketing formation of the tetrahedral intermediate or transition state and can be used to model this hypothetical structure, as depicted in Fig. 6. Presuming that CEA binds as the multiply charged species shown in Fig. 5 at neutral pH, the critical α-ammonium ion must be deprotonated to initiate the acyl substitution reaction. We propose that this result could be facilitated by a catalytic dyad consisting of Tyr-348 and Glu-382 (indicated as B in Fig. 5). In all of the β-LS structures, there is a strong hydrogen bond between the hydroxyl group of Tyr-348 and a side chain oxygen atom of Glu-382. Moreover, in the CMA-AMP/PP_i/Mg²⁺ structure, the Tyr-348 hydroxyl oxygen atom is ≈3.5 Å from the CMA α-nitrogen (Fig. 4). It is reasonable that this residue could move closer to the substrate during β-lactam formation, especially because significant conformational flexibility is observed in the different structures (Fig. 4). In the CEA⋅AMP-CPP/Mg²⁺ and DGPC⋅AMP/PP_i/Mg²⁺ structures, Tyr-348 is further away, with a hydroxyl oxygen to α-nitrogen distance of ≈5 Å. Notably, both Tyr-348 and Glu-382 are conserved in CarA, a β-LS homolog that catalyzes similar chemistry in carbapenem biosynthesis (29, 30). Neither residue is conserved in AS-B, although Glu-382 is replaced with aspartic acid.

Fig 6.

Schematic diagram of the proposed oxyanion intermediate or transition state modeled into the β-LS active site.

The two Mg²⁺ ions also play critical roles in β-lactam formation, neutralizing the negative charge on the PP_i and nascent AMP. Together, they exert a coordinative “pull” effect on the adenylate as the nitrogen nucleophile attacks the sp² center of the adenylated CEA β-carboxylate. Given the extent of substrate preorganization in the active site, we favor formation of a tetrahedral intermediate or transition state, which is apparently stabilized by Lys-443, the ζ nitrogen of which is ≈3.4 Å from CMA adenylate β-carboxylate (Fig. 4). In the DGPC⋅AMP/PP_i/Mg²⁺ structure, the Lys-443 ζ nitrogen is within 3 Å of the DGPC β-lactam carbonyl oxygen atom and would presumably be even closer to the oxygen in the oxyanion. The loop comprising residues 444–453 prevents the intermediate from diffusing out of the active site and protects it from hydrolysis. In the decomposition of the tetrahedral transition state or intermediate, loss of AMP could be synchronous with deprotonation of the α-ammonium ion (pK_a ≈ 9; ref. 31), or take place in a more stepwise manner with AMP as the nearest base to accept the proton from the developing β-lactam nitrogen (pK_a ≈ 0, Fig. 6). In the DGPC⋅AMP/PP_i/Mg²⁺ structure, the distance between the β-lactam nitrogen and the closest AMP phosphate oxygen is ≈3.4 Å, although a water molecule or other acceptor is possible. In both the CMA-AMP/PP_i/Mg²⁺ and the DGPC⋅AMP/PP_i/Mg²⁺ structures, PP_i is located deep in the active site cleft covered by the other ligands and residues 444–453. This observation is consistent with kinetic studies suggesting that PP_i is the last product released (11).

Taken together, these structures encompass the entire β-LS catalytic cycle and allow visualization of each reaction step in sequence, although the detailed mechanism of β-lactam formation remains to be elucidated. Perhaps the most striking observation is the presence of two Mg²⁺ ions, which are crucial to the adenylation and β-lactamization reactions. The structural data also provide new insight into the evolutionary relationship between β-LS and AS-B. Besides the previously observed differences in the N-terminal domains and active site dimensions (10), the structures suggest specific details of active site optimization for β-lactamization that are subject to experimental test. Understanding the molecular features that allow one catalytic activity to evolve into another will form the foundation of future biosynthetic pathway engineering experiments.

Abbreviations

• β-LS, β-lactam synthetase

• AS-B, class B asparagine synthetase

• CEA, N²-(carboxyethyl)-l-arginine

• AMP-CPP, α,β-methyleneadenosine 5′-triphosphate

• CMA, N²-(carboxymethyl)-l-arginine

• DGPC, deoxyguanidinoproclavaminic acid