The primary role of metabolic chemistry is to maintain the low entropy state of the cell. The discovery of nonenzymatic reactions similar to those found in metabolism based on the identity of substrates and products is useful in that such work indicates what types of chemistry were potentially accessible on the prebiotic Earth.¹ However, such prebiotic chemistry would also have been inconsequential if not coupled to the thermodynamically unfavorable reactions needed to sustain the cell. Extant biology does this, in large part, by plugging into oxidative chemistry through the electron shuttling activity of nicotinamide adenine dinucleotide. To date, no examples of prebiotic, metabolic-like chemistry have demonstrated the capturing of electrons with prebiotically plausible electron shuttling molecules.

Despite the considerable focus placed on the RNA world,²⁻⁴ little attention has been given to the redox active dinucleotides found in biology. Fragments of the prebiotic synthesis of NAD⁺ (1) were reported,⁵⁻⁷ and 1 can engage in electron transfer with iron–sulfur peptides,⁸ rhodium trisbipyridine complexes,^9,10 and titanium dioxide particles.¹¹ Further, a laboratory evolved ribozyme can couple the oxidation of a benzyl alcohol to the reduction of 1 to NADH (2).¹²

We sought to determine whether NAD⁺ (1) could have engaged in catabolic-like reactions that could have ultimately driven cell-like activity (Scheme 1). Here, we show that 1 can acquire electrons from α-keto acids, including pyruvate, during oxidative decarboxylation. Several prebiotic routes for the synthesis of pyruvate have been reported.¹³⁻¹⁵ Subsequently, the produced NADH (2) nonenzymatically reduces FAD to FADH₂. Finally, the electrons are deposited into a model terminal electron acceptor, methylene blue. Similar electron transport chains may have impacted the prebiotic chemistry that became metabolism.

Open in a separate window

Scheme 1

Electron Transfer from α-Keto Acids

When NAD⁺ (1) was incubated with pyruvate (3) at 23 °C at pH 9.0 for 24 h, ¹H NMR spectra showed a decrease in the resonances of 1 and an increase in resonances that corresponded to acetate (4) and NADH (2) (Figures ​Figures11a and S1). Integration of the peaks revealed that the starting concentrations of 1 (15 mM) and 3 (30 mM) decreased to 8.1 and 19.7 mM, respectively. After 24 h, 0.9 mM 4 and 1.0 mM 2 formed, indicating that <10% of the consumed 3 and 1 led to the formation of 4 and 2. Pyruvate (3) was not consumed under the same experimental conditions in the absence of 1 (Figure S2). Despite the inefficiency of the reaction, reduction of 1 to 2 strongly correlated with the production of 4 (ρ = 0.95) (Figure S3), consistent with a dependence on the oxidative decarboxylation of 3. The presence of 4 was confirmed by 1D and 2D heteronuclear NMR with ¹³C labeled 3 (Figures S4–S6). The production of 2 was confirmed by UV–vis absorption and fluorescence spectrophotometry (Figure S7). After 1 h, an absorption band at 340 nm and an emission band at 450 nm emerged that were consistent with the presence of 2. After 24 h, an additional absorption band appeared at 430 nm, and the emission peak shifted to 505 nm, suggesting the formation of degradation products derived from 1. LC–MS of these samples (Figure S8) corroborated the production of 2 (m/z = 664.1168, C₂₁H₂₉N₇O₁₄P₂–H^–) after 1 and 24 h. Since NADH (2) can exist in two different major isomeric forms (1,4-NADH and 1,6-NADH) that can be difficult to distinguish, the identity of 2 was verified by the enzymatic activity of lactate dehydrogenase,¹⁰ which only mediates the oxidation of 1,4-NADH. After incubation of 1 and 3 for 24 h, the addition of lactate dehydrogenase led to an immediate loss of the resonances of the produced 2 (Figure S9). Taken together, the data were consistent with the nonenzymatic reduction of 1 to 2 by 3.

Open in a separate window

Figure 1

(a) ¹H NMR spectra of the reaction products from incubation of 1 and 3 showed formation of 2 and 4. (b) The reduction of 1 (15 mM) to 2 with 3 (24 h). 3 was at 1.5, 7.5, 15, 22.5, 30, and 150 mM. Data are mean and SEM (standard error of the mean) of distinct samples, n = 2. (c) HPLC chromatograms show the formation of 2 after 24 h. All reactions were in 200 mM bicarbonate, pH 9.0.

The formation of NADH (2) was dependent on the concentration of pyruvate (3) (Figures ​Figures11b and S10) with a Pearson correlation coefficient of 0.94. A 150 mM concentration of 3 reduced 13.4 ± 0.6% of 15 mM NAD⁺ (1) to NADH (2) after 24 h. Lower concentrations of 3 reduced less 1 under the same experimental conditions (Figure ​Figure11b). For example, 30 mM 3 led to the reduction of 6.5 ± 0.4% of 1 to 2. Similarly, titrations of 1 with a constant concentration of 3 showed a direct correlation between 1 and 3 (ρ = 0.99), as expected (Figure S11).

Reduction of NAD⁺ (1) to NADH (2) by pyruvate (3) was dependent on pH (Figure S12). 2 was not produced below pH 8. Since the reactions at pH 7.5 and 8.0 were in 200 mM sodium phosphate and the reactions at higher pH were in 200 mM bicarbonate, it was not clear if the increased reactivity at higher pH was due to pH or participation of the buffer. Therefore, 1 was incubated with 3 in 200 mM pyrophosphate, pH 9.0, which led to the reduction of 5.2 ± 0.9% of 1 to 2 (Figure S14). The data were within error of the 6.5 ± 0.4% observed in 200 mM bicarbonate. Thus, the buffer did not play a significant role. At pH above 7.0, some degradation of 1 was observed (Figure S15a), with 26.0 ± 3.5% of 1 degraded at pH 9 after 24 h. Conversely, no degradation of 3 was detected by ¹H NMR spectroscopy over the same pH range. Further, no differences in the equilibrium between 3, hydrated-3, and the formation of parapyruvate were detected (Figure S15b). Reactions at higher temperatures were not pursued because of the increased degradation of 1 (Figure S16). Ambient light did not facilitate the reduction of 1 (Figure S16).¹⁶

Since α-keto acids undergo facile decarboxylation,¹⁷⁻¹⁹ the ability of α-keto acids other than pyruvate (3) to reduce NAD⁺ (1) was investigated. The incubation of 1 with oxaloacetate (5), α-ketoglutarate (6), 2-ketobutyrate (7), and 2-oxooctanoate (8) all gave rise to NADH (2) (Figures ​Figures22 and S17). However, the yield of 2 was greater with 3 and 7. For example, 74% less 2 was produced at pH 9.0 with 5 than with 3. All reactions with α-keto acids showed increased activity at higher pH (Figure S12). The ¹H NMR spectra of reactions with 5 were similar to reactions with 3, because 5 decarboxylates to 3 (Figure S18).²⁰ Decreased yields of 2 with 5 and 6 suggest that the additional carboxylate interferes with the mechanism, perhaps by raising the pK_a of the enol. The decreased reactivity of 8 likely stemmed from the formation of aggregates.²¹

Open in a separate window

Figure 2

(a) Chemical structures of tested α-keto acids. (b) Reduction of 1 to 2 by α-keto acids. A 30 mM concentration of each α-keto acid (3 and 5–8) was incubated with 1 (15 mM) for 24 h at pH 9.0. Analysis was by HPLC. Data are mean and SEM of distinct samples, n = 2.

Next, we investigated the importance of the α-carbonyl by testing the activity of molecules with a β-carbonyl (acetoacetic acid, 9) and a γ-carbonyl (levulinic acid, 10) (Figures S19 and S20). Both molecules failed to reduce detectable amounts of NAD⁺ (1) as judged by ¹H NMR spectroscopy. The activity of glyoxylate (11), which contains an α-aldehyde as opposed to an α-ketone, was also evaluated. Glyoxylate (11) was unable to reduce 1 to 2 (Figure S21). Finally, other components of the citric acid cycle that enzymatically provide electrons to 1, such as malate (12) and isocitrate (13), were tested. Neither 12 nor 13 were capable of reducing 1 to 2 (Figure S22). Of the molecules tested, only enolizable α-keto acids were found to nonenzymatically reduce 1 to 2, possibly through a single electron transfer mechanism (Figures S23 and 24), although other mechanisms are possible.

In addition to NAD⁺ (1), FAD (14) is a redox active ribodinucleotide that may have played a role in early protocellular systems (Figure S25).²² In extant biology, NADH (2) frequently reduces flavins, e.g., NADH dehydrogenase. Consistent with previous reports of the nonenzymatic reduction of 14 by 2,¹² the electrons from pyruvate (3) were shuttled to 14 via NAD(H) (Figure ​Figure33a). More specifically, if 3 was premixed and incubated with 1 for 24 h, the ¹H resonances of 2 diminished to undetectable levels upon the addition of 14 (Figure S26). If all three, i.e., 1, 3, and 14, were incubated together from the beginning, ¹H NMR spectra showed a clear dependence on the amount of FADH₂ (15) produced on the starting concentration of 1 (Figure S27). UV–vis absorption spectroscopy corroborated the reduction of 14 to 15 (Figures ​Figures33a and S28). 14 absorbs strongly, whereas 15 is colorless.²³ The reaction was complete within 8 h, and no reduction of 14 was observed in the absence of 3 or 1. Similarly, 14 was not reduced if 3 was replaced with a nonactive electron donor, e.g., the β-keto acid 9 (Figure ​Figure33a).

Open in a separate window

Figure 3

(a) Assembly of an electron transport chain. (b) The reduction of 14 (●). The reaction contained 30 mM 3, 15 mM 1, 0.1 mM 14, and 200 mM bicarbonate, pH 9.0, and was monitored at 475 nm. The reaction did not proceed in the absence of 1 (○) or 3 () or in the presence of the nonactive electron donor 9 (□). (c) The transfer of electrons from 3 to 16 (●). Reactions contained 30 mM 3, 15 mM 1, 7.5 mM 14, 13.3 μM 16, and 200 mM bicarbonate, pH 9.0. 16 was not reduced in the absence of either 1 (), 3 (○), or 14 (□). Data are mean and SEM of distinct samples, n = 2.

The electron transport chain was completed with a model terminal electron acceptor, methylene blue (16). Consistent with the electron shuttling roles of NAD⁺ (1) and FAD (14), the electrons from pyruvate (3) ultimately reduced 16 (Figures ​Figures33b and S29) after incubation for 120 min. 16 was not reduced when either 3, 1, or 14 were not included in the reaction mixture. A more prebiotically plausible terminal electron acceptor would be ferricyanide.²⁴ However, the use of ferricyanide would require spatial separation, since ferricyanide can oxidize 2 and 15. Ferricyanide did not oxidize 3 under the exploited conditions (Figure S30).

In the modern-day citric acid cycle, one α-keto acid (α-ketoglutarate, (6)) and two α-hydroxy acids (malate (12) and isocitrate (13)) are oxidized in a manner that reduces NAD⁺ (1) to NADH (2).^25,26 Of these reactions, only 6 was found to reduce NAD⁺ (1) nonenzymatically, albeit inefficiently. However, oxaloacetate (5) was able to reduce 1 nonenzymatically, even if 5 is not oxidatively decarboxylated, and 5 does not reduce 1 in the contemporary citric acid cycle. Therefore, if a prebiotic, NAD⁺-dependent metabolic-like pathway existed, this pathway would likely have been different from what is observed in biology today. One possibility is that prebiotic analogues of contemporary metabolism more fully dependent on α-keto acids.^27,28 For example, a nonenzymatic α-keto acid version of the citric acid cycle was recently reported.²⁸ Although this cycle runs in reverse, using glyoxylate to reduce α-keto acids, each of the α-keto acid intermediates would be capable of reducing 1. Therefore, even if glyoxylate does not reduce NAD⁺ directly, the electrons from glyoxylate could be transferred through α-keto acids to NAD⁺, thus initiating a thermodynamically favorable electron transport chain. Another possibility is that NAD⁺ could be reduced by 5 through the activity of a prebiotic, α-keto acid-dependent glyoxylate cycle.²⁷ This cycle was reported to proceed via the oxidative decarboxylation of 5 with hydrogen peroxide. 1 can mediate the same reaction without the loss of electrons. Further efforts in combining the redox active components of the RNA world with oxidative, metabolic-like chemistry will likely give critical insights into the neglected but necessary nonenzymatic steps between abiotic and biotic chemistry.