arising from A. Edwards et al. Nature Communications https://doi.org/10.1038/s41467-023-36503-2 (2023)

Grass pea (Lathyrus sativus L.) is a nutritious and robust grain legume grown mainly in Asia and Africa that also produces a neurotoxic compound, β-N-oxalyl-l-α,β-diaminopropionic acid (β-ODAP). Recently, Edwards et al. 1 published an annotated genome assembly of grass pea and proposed that β-ODAP is synthesized by the enzyme LsAAE3, and is stimulated by binding to coenzyme A (CoA) and the enzyme LsBOS. In contrast, we had previously identified and characterized LsAAE3 as an oxalyl-CoA synthetase (LsOCS) from grass pea2, and LsBOS as the enzyme that synthesizes β-ODAP in grass pea by ligating the metabolites oxalyl-CoA and L-α,β-diaminopropionic acid (L-DAPA)3. The importance of this issue stems from the need to curtail β-ODAP production in grass pea by genetic engineering in order to detoxify it, enabling its safe consumption as food and fodder.

Unlike our previous findings3, Edwards et al. stated that the in vitro incubation of recombinant, purified LsBOS with oxalyl-CoA and L-DAPA “did not produce any β-ODAP.” Instead, they claimed that β-ODAP is produced from oxalic acid and L-DAPA together with ATP and Mg2+ by the oxalyl-CoA synthetase LsAAE3 at physiological pH 7.5. They suggested that oxalyl-AMP, which is initially synthesized in LsAAE3 by coupling ATP and oxalic acid, is then ligated directly to L-DAPA by LsAAE3 to form the product β-ODAP. The role of LsBOS, according to Edwards et al., is to direct the regio-selectivity of the reaction, which can produce both α- and β-ODAP by LsAAE3 on its own, by an unknown mechanism. Additionally, they stated that the complete reaction is stimulated by nanomolar concentrations of CoA, also by an unknown mechanism1.

Edwards et al. examined the kinetic parameters of LsAAE3 and showed it could produce oxalyl-CoA from oxalic acid, ATP, and Mg2+. However, they claimed this activity is optimal for LsAAE3 at pH 6 and that “virtually no activity” remains at pH 7.0 and above. In contrast, we previously showed that LsAAE3 has an optimal pH of 8.0, similar to the ones reported for its close homologs AtAAE34, MtAAE35, GsAAE36, and ScAAE37 and that this activity is reduced ~5 fold at pH 6.02. The authors reported the high sequence identity of LsAAE3 to its homologs from A. thaliana (75%) and M. truncatula (88%) and indicated that the normal environmental pH in the cell cytoplasm where LsAAE3 is thought to be located is 7.5. Yet, they determined an optimal catalytic efficiency for oxalyl-CoA synthesis at pH 6 using LC-MS and found it to be 7.5 × 10−2 s−1M−1, a value that is 7 orders of magnitude lower than the one we reported for the same enzyme2, 1.1 × 105 s−1M−1, at pH 8. The value we found is also similar to the ones reported by others for oxalyl-CoA synthesis by its close homologs AtAAE34, MtAAE35, GsAAE36, and ScAAE37, all at optimal physiological pHs of 7.5−8.

This large discrepancy in the catalytic efficiencies of LsAAE3 as an oxalyl-CoA synthetase stems from a very low kcat value of 1.5 × 10−4 s−1 reported by Edwards et al. in comparison to the one we reported (7.6 s−1)2 and a higher KM value for oxalate (2 mM) relative to the one we reported for LsAAE3 (71.5 μM)2 and others reported for homologous AAE3’s (i.e. KM 20−149 μM)4,5,6,7. The low efficiency of oxalyl-CoA synthesis determined by Edwards et al. is surprising in light of the fact that among plants, the highest oxalyl-CoA synthetase activities had been measured in extracts of grass pea, pea, soybean, pumpkin, and wheat germ4. In fact, robust oxalyl-CoA synthetase activity was initially identified in L. sativus extracts and attributed to an oxalyl-CoA synthetase more than 50 years ago8,9. The incorporation of 14C-labeled oxalic acid into oxalyl-CoA was shown to be dependent on ATP, CoA, and Mg2+ at pH 7.5, and the labeled oxalate incorporated into β-ODAP upon the addition of L-DAPA to the plant extract8,9,10.

Contrary to Edwards et al., we previously showed that β-ODAP is produced in vitro from oxalyl-CoA and L-DAPA by LsBOS at physiological pH3. We used purified LsAAE3 to synthesize oxalyl-CoA as a substrate for LsBOS, as we were unable to obtain it commercially. Edwards et al. argued that the in vitro formation of β-ODAP we reported3 resulted from the interaction of LsBOS with residual LsAAE3 in our preps of oxalyl-CoA and could be ascribed to the mechanism they proposed. However, this claim does not explain the activities we observed with LsBOS in reactions not containing LsAAE3 when we coupled L-DAPA to other acyl-CoA substrates such as acetyl-CoA, malonyl-CoA, and glutaryl-CoA that were obtained commercially as pure compounds3.

Even so, we decided to examine this proposal by eliminating any residual LsAAE3 in our preps and repeating the in vitro assay with LsBOS. For this, we synthesized oxalyl-CoA using LsAAE3, purified it by HPLC (Supplementary Figs. 1 and 2a), and added it to a buffered solution containing an equimolar concentration of L-DAPA. As expected, incubation of purified oxalyl-CoA and L-DAPA with LsBOS resulted in a robust synthesis of β-ODAP, while incubation of these substrates with LsAAE3 or the substrates with no enzyme in either pH 7.5 or pH 9 resulted in only background levels of β-ODAP synthesis (Fig. 1). Thus, the β-ODAP synthesis activity we observed results from the coupling of oxalyl-CoA and L-DAPA catalyzed solely by purified LsBOS.

Fig. 1: Synthesis of β-ODAP from oxalyl-CoA and L-DAPA by LsBOS.
figure 1

Purified oxalyl-CoA (0.4 mM) was incubated with L-DAPA (0.4 mM) and LsBOS (0.15 μM) for 1 h at 37 °C at pH 7.5 (complete reaction). Similar reaction mixtures containing either: no enzyme (No Enz) or with LsAAE3 (0.15 μM) replacing LsBOS (LsAAE3) at pH 7.5 or pH 9 were similarly incubated for 1 h at 37 °C. A control reaction containing only LsAAE3 (0.15μM) and oxalyl-CoA (0.4 mM) at pH 7.5 (LsAAE3 + oxalyl-CoA) was similarly incubated. The reactions were inactivated, diluted 1:10, and the amount of β-ODAP product was quantified in each reaction using LC-MS. Source data are provided as a Source Data file.

Full size image

If L-DAPA is also a substrate of LsAAE3, as Edwards et al. claimed, it should influence the kinetics of the enzyme. To examine the effect of L-DAPA on the rate of oxalyl-CoA synthesis by LsAAE3, we incubated the purified enzyme with ATP, Mg2+, CoA, and oxalic acid with or without L-DAPA and monitored the reaction rate using a coupled NADH enzymatic assay system. We did not observe any significant difference in reaction rates when comparing reactions containing oxalic acid only to reactions in which L-DAPA was added at a 6-fold higher concentration than that of oxalic acid and 1.5-fold than that of CoA (Fig. 2). Previously2, we have shown that under these conditions CoA is fully utilized as a substrate. Thus, L-DAPA does not seem to act as a substrate of LsAAE3.

Fig. 2: Kinetics of oxalate ligation to CoA by LsAAE3 with and without L-DAPA.
figure 2

The rates of CoA (2 mM) ligation to oxalate (0.5 mM) were assayed at pH 7.5 using purified LsAAE3 (1μM) in the presence of ATP (5 mM) and Mg+2 (5 mM; blue bar) and with the addition of L-DAPA (3 mM; red bar). Initial reaction rates were determined using an NADH-coupled enzymatic assay. Error bars denote the SD of four independent replicates (n = 4). Data was analyzed using a two-tailed t-test. ns – not significant (p ≥ 0.05). Source data are provided as a Source Data file.

Full size image

The claims of Edwards et al. regarding the ability of LsAAE3 to synthesize β-ODAP on its own from oxalic acid, CoA, L-DAPA, ATP, and Mg2+ were based on measurements of ATP hydrolysis rates using an NADH-coupled enzymatic reaction and LC-MS identifications of the β-ODAP product. These measurements were made at pH 8.0, at which the chemical reaction between the high-energy oxalyl-CoA substrate and L-DAPA can occur spontaneously, albeit at a much lower rate than that of a reaction enzymatically catalyzed by LsBOS (Fig. 1). In fact, the ability of acyl-CoA thioesters to acylate the ε-amino groups of protein lysine residues, similar in structure to L-DAPA, in a non-enzymatic manner is well known11. In the absence of CoA, Edwards et al. were able to show the formation of α- and β-ODAP in much smaller amounts. Under such conditions, it is likely that LsAAE3 catalyzes the formation of oxalyl-AMP, which in the absence of CoA, is then released to the solution, as previously observed with valproic and cholic acids12,13. This high-energy compound can then react with the amine groups of L-DAPA to form α- or β-ODAP and may explain the identification of these compounds by Edwards. To test this possibility, we incubated LsAAE3 in the presence of oxalic acid, ATP, and Mg2+ and replaced L-DAPA with L-Lysine. The latter has a similar structure to L-DAPA but is not the “cognate” substrate of LsAAE3, according to Edwards et al. Still, we were able to detect the formation of oxalyl-L-Lysine by LC-MS (Supplementary Fig. 3), supporting the idea that primary amine-oxylations can occur by an in vitro chemical reaction with oxalyl-AMP that had been synthesized by LsAAE3 in the absence of CoA.

When CoA is added to a reaction containing oxalic acid, ATP, Mg2+, L-DAPA, and both LsAAE3 and LsBOS, it can produce oxalyl-CoA, but the CoA will be regenerated following the reaction with L-DAPA. It is thus not surprising that in such a mixture, low concentrations of CoA are required for measuring both ATP hydrolysis and the formation of β-ODAP. It does not confirm, however, the mechanism suggested by Edwards et al.

Edwards et al. provided in vitro evidence for binding between LsAAE3 and LsBOS and claimed that this interaction “enables regio-selective amide formation” in LsAAE3, yet they did not provide any explanation as to the mechanism. Their suggestion does not explain why we could not observe any β-ODAP synthesis when we incubated oxalyl-CoA and L-DAPA with LsBOS variants that we mutated in either one of its two catalytic residues, His162 and Asp166, located deep inside its active site3. These single residue LsBOS mutants were expressed and purified as soluble but inactive proteins supporting their essential roles in the catalytic mechanism we proposed for LsBOS and not the auxiliary regulatory role proposed by Edwards et al.

To rule out the production of oxalyl-CoA as a precursor metabolite required for β-ODAP synthesis, Edwards et al. suggested that LsAAE3 had undergone an “evolutionary repurposing” of its activity as an oxalyl-CoA synthetase to a β-ODAP synthetase. However, to explain how the heterologous expression of LsBOS in the leaves of N. benthamiana together with the introduction of L-DAPA resulted in the production of β-ODAP in both of our reports1,3, they attributed a similar mechanism of β-ODAP synthesis to the homolog of LsAAE3 in tobacco (NbAAE3); i.e., that NbAAE3, could also interact with L-DAPA and LsBOS to form β-L-ODAP. In contrast to the high abundance of oxalate and oxalyl-CoA synthesizing enzymes in plants such as N. benthamiana, β-ODAP is a unique metabolite found only in a number of species of L. sativus, Crotalaria, Acacia, and Panax. We find the idea that the same specialized interaction between a unique enzyme, LsBOS, and its evolutionary adapted binding partner, LsAAE3, can be readily replicated in a heterologous host that does not contain a homologous β-ODAP synthase unlikely. Moreover, we introduced L-DAPA by itself to N. benthamiana and found no evidence of any α- or β-ODAP production3. LsAAE3 and LsBOS may have, indeed, co-evolved in L. sativus to bind and interact in their native cellular environment. Yet this adaptation had likely resulted from the need to improve the metabolic flux of β-ODAP production in grass pea by providing LsBOS with greater access to its substrate oxalyl-CoA.

Finally, we examined the chemically synthesized oxalyl-CoA that was used in the work of Edwards et al.1 and was kindly provided by them. When incubated with equimolar concentrations of L-DAPA and our purified enzyme LsBOS, we were unable to detect the formation of β-ODAP or the consumption of L-DAPA (Supplementary Fig. 4). In contrast, L-DAPA was completely consumed and β-ODAP was produced in control reactions that used enzymatically synthesized oxalyl-CoA (Supplementary Fig. 4). When we examined the compound of Edwards et al. by UPLC-MS (Supplementary Fig. 2b) and NMR (Supplementary Figs. 5–11) we found that it did not contain oxalyl-CoA, but instead contained a derivative of propanoyl-CoA. This may explain the inability of Edwards et al. to detect the formation of β-ODAP by LsBOS in similar in vitro reactions1.

To conclude, our past results and those presented herein support a two-step mechanism in which LsAAE3 first synthesizes oxalyl-CoA from oxalic acid, ATP, and CoA. Oxalyl-CoA is then taken up as a substrate together with L-DAPA by LsBOS, which catalyzes the production of β-ODAP in grass pea. The two enzymes may, indeed, bind to improve metabolic flux and prevent oxalyl-CoA from interacting with other molecules in the cell. Establishing the correct mechanism of β-ODAP production in grass pea is a prerequisite for effectively employing genetic engineering and additional methods to eliminate its production in grass pea cultivars. Obtaining cultivars devoid of β-ODAP will increase the safety and use of grass pea as food and fodder.

Methods

In vitro synthesis and purification of oxalyl-CoA

The gene encoding oxalyl CoA-synthase (LsAAE3) was amplified from extracted genomic DNA of L. sativus, cloned into a bacterial expression vector, expressed in E. coli, and purified2. Purified LsAAE3 (0.25 μM) was added to a reaction solution (125 mM Tris, pH 7.5; 5 mM MgCl2; 5 mM ATP) containing oxalic acid (3 mM) and coenzyme A (1.5 mM) and incubated at 37 °C for 1 h. The reaction solution was filtered through 0.2 μm PTFE filters (StarTech®) prior to HPLC purification and separated by reversed-phased chromatography on a C18 column (Gemini C18, 4.6×150 mm, 5 μM, Phenomenex©). The column was attached to a Prominence UFLC LC-20AD system (Shimadzu©) consisting of a SIL-20AC autosampler (Shimadzu©), CTO-20AC column oven (Shimadzu©) and a SPD-M20A diode array detector (Shimadzu©). Elution was done using a gradient of 0-50% solvent B (100% acetonitrile) in solvent A (13 mM sodium phosphate buffer, pH 7.4) at a flow rate of 1 mL per minute for 20 minutes at 25 °C while monitoring at 260 nm. Data analysis was performed using LabSolutions ver. 5.97 (Shimadzu©). Fractions of 600 μL were collected and purified oxalyl-CoA was identified by LC-MS using a Xevo G2-S QTof high-resolution mass spectrometer (HRMS) coupled with an Acquity I-class UPLC (Waters) as follows: 5 μL of purified oxalyl-CoA (0.4 mM) or 5 μL of CoA (10 mM) control sample were injected onto a BEH Amide column (Acquity, 1.7 μm, 2.1 × 150 mm, Waters) using a gradient of 10 mM ammonium carbonate, pH 10 in acetonitrile. Mass spectrometry was performed using an MS scan in the 50–1600 m/z range (negative mode) by analysis of total ion current or extracted ion chromatograms. CoA was identified using the 766 m/z peak and oxalyl-CoA using the 838 m/z peak as [M-H]- ions. The compound provided by Edwards et al. was analyzed in the same manner.

In vitro synthesis reactions of β-ODAP

Oxalyl-CoA (0.4 mM) was mixed with L-DAPA (0.4 mM), and either: LsBOS3 (0.15 μM) or LsAAE3 (0.15 µM) or no enzyme in buffer (50 mM Tris, pH 7.5) or (50 mM Tris, pH 9) and incubated for 1 h at 37 °C. Following inactivation at 65 °C, samples were analyzed for β-ODAP using LC-MS/MS as previously described3.

Kinetics of oxalyl-CoA synthesis

The enzymatic activity of purified LsAAE3 was assayed using a coupled enzymatic assay as described2. Briefly, the assay mixture contained: HEPES buffer (50 mM, pH 7.5), MgCl2 (5 mM), phosphoenol-pyruvate (3 mM), NADH (1 mM), ATP (5 mM), DTT (2 mM), CoA (2 mM), purified LsAAE3 (1 μM), myokinase (5U), pyruvate kinase (5U), lactate dehydrogenase (6U) with or without L-DAPA (3 mM). The reaction was initiated by the addition of the buffered oxalate solution (0.5 mM, pH 7.5) to the reaction mix containing the LsAAE3 enzyme and above described compounds in UV-transparent 96-well plates (final volume of 200 μL) at room temperature. The rate of NADH oxidation was then determined by measuring absorbance at 340 nm for 20 min using an Epoch2 microplate reader (Biotek©).

NMR analysis of compounds

All NMR experiments were recorded on an 11.75T (500.08 MHz) Bruker AVANCE III HD spectrometer at 298 K. The chemical shifts (δ) are given in ppm relative to a residual solvent resonance (4.7 ppm for D2O). The 1H DOSY NMR measurement was performed on an NMR spectrometer equipped with a 50 gauss/cm Z gradient system. The LED (longitudinal eddy current delay) pulse sequence was used with smoothed square (SMSQ.10.100) gradients. The gradients were incremented from 2% to 98% in 32 linear steps and 24 scans were acquired for each gradient. The gradient duration was 2 ms and the diffusion time was 50 ms. The recycle delay was 3 s resulting in a total experiment time of 57 min. The 1H-1H COSY experiment was performed with a 2Kx256 matrix, zero filled to 2Kx512, with 8 scans and a recycle delay of 2 s. The 1H-1H NOESY experiment was performed with the same experimental parameters as the COSY experiment with a mixing time of 0.5 s. The 1H-13C HSQC and 1H-13C HMBC experiments were acquired with a spectral width of 16ppm for 1H and 230 ppm for the 13C, a matrix of 2Kx512 (zero filled to 2Kx1K), 2 s recycle delay and with 8 and 12 scans, respectively.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.