The Reaction of Aminonitriles with Aminothiols: A Way to Thiol-Containing Peptides and Nitrogen Heterocycles in the Primitive Earth Ocean

The Strecker reaction of aldehydes with ammonia and hydrogen cyanide first leads to α-aminonitriles, which are then hydrolyzed to α-amino acids. However, before reacting with water, these aminonitriles can be trapped by aminothiols, such as cysteine or homocysteine, to give 5- or 6-membered ring heterocycles, which in turn are hydrolyzed to dipeptides. We propose that this two-step process enabled the formation of thiol-containing dipeptides in the primitive ocean. These small peptides are able to promote the formation of other peptide bonds and of heterocyclic molecules. Theoretical calculations support our experimental results. They predict that α-aminonitriles should be more reactive than other nitriles, and that imidazoles should be formed from transiently formed amidinonitriles. Overall, this set of reactions delineates a possible early stage of the development of organic chemistry, hence of life, on Earth dominated by nitriles and thiol-rich peptides (TRP).


Introduction
In ribosomes, peptide bonds are formed by the reaction of the amine group of an amino acid with an ester function. For non-ribosomal peptides, the amide formation involves the reaction of an amine on a thioester [1]. In both cases, mixed phosphoric carboxylic anhydrides are transiently formed. Esters, thioesters, and anhydrides are activated forms of the carboxylic acid function. Their intermediacy is mandatory and no significant C-N bond formation would occur directly from the reaction of an acid function with an amine [2]. What is true in today's biology, was also true four billion years ago, when life was beginning its development in the terrestrial ocean. Activated derivatives had to be involved in the formation of prebiotic polymers. As a consequence, if acids were involved at some stage, a strong energy source was necessary. Nowadays, it is furnished by the cleavage of the triphosphate group of adenosine triphosphate [3].
Many simple aldehydes were probably present in the primitive ocean [4] and are plausible precursors for α-amino acids. Reacting with ammonia and hydrogen cyanide, they would have first given α-aminonitriles, which, upon hydrolysis, would have delivered amino acids ( Figure 1) [5]. However, even though it is exothermic, the reaction of nitriles with water is a slow process [6]; so slow that, once formed in the ocean, aminonitriles would have had ample time to react with species more nucleophilic than water.  Nitriles are known to react with aminothiols to give thiazolines, which in turn can be hydrolyzed to mercaptoamides [7]. Starting from α-aminonitriles and cysteine, the expected products of this twostep process are dipeptides ( Figure 1). In the early ocean, this could have been an efficient and selective process to thiol-containing dipeptides [8].
Compared to any activation process starting from acids, this nitrile scenario has the advantage of not necessitating any strong energy source. The acid does not need to be activated as it is delivered directly in an activated form by the Strecker reaction.
HCN has long been given an important role in prebiotic molecular evolution [9]. As it is largely distributed in space, having been observed in various regions, for instance, near carbon stars [10] and in a proto-planetary nebula [11], as well as in comets [12,13], it is highly possible that HCN was present on the early Earth. Furthermore, it has been postulated that it could have been formed when numerous asteroids struck our planet during the Late Heavy Bombardment [14]. It might have been produced photochemically in the atmosphere [15,16]. It was ejected from volcanoes [17] and submarine hydrothermal vents [18].
Hydrogen sulfide is another important small molecule in our hypothesis. It would have been necessary for the formation of cysteine. It has often been detected in space [19], inter alia in star forming regions [20], and in cold clouds [21], as well as in comets [22]. Furthermore, it is abundantly ejected from volcanoes [23,24], so there is no doubt that it was effectively present on the primitive Earth. Its presence permitted the synthesis of cysteine and homocysteine [25]. Homocysteine would have been obtained by a Strecker reaction starting from the addition of the product H2S onto acrolein (HSCH2CH2CHO). In a similar way, cysteine would have been synthesized from HSCH2CHO, itself possibly obtained from glycolaldehyde.

Experimental Section
Products (thiazolines, dipeptides…) were identified in reaction mixtures by NMR spectroscopy ( 1 H and 13 C) and mass spectrometry. No attempt at purifying them was made (except for 11 and 12).
For the mass experiment, H2O was used as the solvent. High-resolution mass spectra were recorded on a Waters G2-S Q-TOF mass spectrometer or on a LTQ Orbitrap XL (Thermo Scientific) spectrometer. Low resolution ESI analysis was performed on an Amazon speed (Brucker Daltonics) IonTrap spectrometer.
Compared to any activation process starting from acids, this nitrile scenario has the advantage of not necessitating any strong energy source. The acid does not need to be activated as it is delivered directly in an activated form by the Strecker reaction.
HCN has long been given an important role in prebiotic molecular evolution [9]. As it is largely distributed in space, having been observed in various regions, for instance, near carbon stars [10] and in a proto-planetary nebula [11], as well as in comets [12,13], it is highly possible that HCN was present on the early Earth. Furthermore, it has been postulated that it could have been formed when numerous asteroids struck our planet during the Late Heavy Bombardment [14]. It might have been produced photochemically in the atmosphere [15,16]. It was ejected from volcanoes [17] and submarine hydrothermal vents [18].
Hydrogen sulfide is another important small molecule in our hypothesis. It would have been necessary for the formation of cysteine. It has often been detected in space [19], inter alia in star forming regions [20], and in cold clouds [21], as well as in comets [22]. Furthermore, it is abundantly ejected from volcanoes [23,24], so there is no doubt that it was effectively present on the primitive Earth. Its presence permitted the synthesis of cysteine and homocysteine [25]. Homocysteine would have been obtained by a Strecker reaction starting from the addition of the product H 2 S onto acrolein (HSCH 2 CH 2 CHO). In a similar way, cysteine would have been synthesized from HSCH 2 CHO, itself possibly obtained from glycolaldehyde.

Experimental Section
Products (thiazolines, dipeptides . . . ) were identified in reaction mixtures by NMR spectroscopy ( 1 H and 13 C) and mass spectrometry. No attempt at purifying them was made (except for 11 and 12).
For the mass experiment, H 2 O was used as the solvent. High-resolution mass spectra were recorded on a Waters G2-S Q-TOF mass spectrometer or on a LTQ Orbitrap XL (Thermo Scientific) spectrometer. Low resolution ESI analysis was performed on an Amazon speed (Brucker Daltonics) IonTrap spectrometer. Theoretical calculations were carried out using the Gaussian09, Revision D.01 software. All the geometries were optimized using the B3LYP functional in conjunction with the 6-31g(d,p) basis set and the water solvent effects were described by using the polarizable continuum model (PCM), namely IEFPCM (integral equation formalism PCM) [26]. These optimizations were followed by a frequency calculation at the same level to ensure that the geometry was indeed a real minimum, i.e., all the second derivatives were positive.

Results
We first studied the reaction of aminoacetonitrile (GlyCN 1a, the nitrile derivative of glycine) with cysteine. Reactions were conducted in D2O solutions and followed by NMR spectroscopy. Representative 1 H NMR spectra are shown in Figure 2. Formation of the expected thiazoline ring 2a was evidenced by the apparition of signals at ca. 5 ppm (a triplet-like dd), and from 3.4 to 3.7 ppm (2 dd). After some time, new signals grew, including a triplet at 4.4 ppm and a thin doublet-like signal at ca. 2.9 ppm, both characteristic of Gly-Cys 3a. We repeated this experiment many times, generally at a concentration of 5 × 10 −3 to 5 × 10 −2 mol/L, for practical NMR measurements. However, we also tested it at 3 10 −4 mol/L, a concentration at which 2a and 3a were also obtained.
Met-CN (168 mg, 1.29 mmol) was dissolved in 15 mL H 2 O. Cys-NH 2 .TFA (280 mg; 1.29 mmol) was added. The pH of the solution was adjusted to 8 by adding Na 2 CO 3 . The solution was stirred at 45 • C for 2.5 h. The aqueous phase was extracted three times with ethyl acetate. The organic layer was dried over Na 2 SO 4 , filtered, and concentrated under vacuum. After purification by silica gel chromatography (1-10% MeOH/DCM), the thiazoline 11 was obtained as an orange oil (16% yield). Theoretical calculations were carried out using the Gaussian09, Revision D.01 software. All the geometries were optimized using the B3LYP functional in conjunction with the 6-31g(d,p) basis set and the water solvent effects were described by using the polarizable continuum model (PCM), namely IEFPCM (integral equation formalism PCM) [26]. These optimizations were followed by a frequency calculation at the same level to ensure that the geometry was indeed a real minimum, i.e., all the second derivatives were positive.

Results
We first studied the reaction of aminoacetonitrile (GlyCN 1a, the nitrile derivative of glycine) with cysteine. Reactions were conducted in D2O solutions and followed by NMR spectroscopy. Representative 1 H NMR spectra are shown in Figure 2. Formation of the expected thiazoline ring 2a was evidenced by the apparition of signals at ca. 5 ppm (a triplet-like dd), and from 3.4 to 3.7 ppm (2 dd). After some time, new signals grew, including a triplet at 4.4 ppm and a thin doublet-like signal at ca. 2.9 ppm, both characteristic of Gly-Cys 3a. We repeated this experiment many times, generally at a concentration of 5 × 10 −3 to 5 × 10 −2 mol/L, for practical NMR measurements. However, we also tested it at 3 10 −4 mol/L, a concentration at which 2a and 3a were also obtained.
Val-CN.HCl (35 mg, 0.26 mmol) was dissolved in 5 mL H 2 O. Cys-NH 2 .TFA (57 mg; 0.26 mmol) was added. The solution was adjusted to pH = 7 by adding Na 2 CO 3 . The solution was stirred at 45 • C for 24 h. The aqueous phase was extracted three times with ethyl acetate. The organic layer was dried over Na 2 SO 4 , filtered, and concentrated under vacuum. After purification by silica gel chromatography (1-10% MeOH/DCM), the thiazoline 12 was obtained as a yellow oil (30% yield). Theoretical calculations were carried out using the Gaussian09, Revision D.01 software. All the geometries were optimized using the B3LYP functional in conjunction with the 6-31g(d,p) basis set and the water solvent effects were described by using the polarizable continuum model (PCM), namely IEFPCM (integral equation formalism PCM) [26]. These optimizations were followed by a frequency calculation at the same level to ensure that the geometry was indeed a real minimum, i.e., all the second derivatives were positive.

Results
We first studied the reaction of aminoacetonitrile (GlyCN 1a, the nitrile derivative of glycine) with cysteine. Reactions were conducted in D 2 O solutions and followed by NMR spectroscopy. Representative 1 H NMR spectra are shown in Figure 2. Formation of the expected thiazoline ring 2a was evidenced by the apparition of signals at ca. 5 ppm (a triplet-like dd), and from 3.4 to 3.7 ppm (2 dd). After some time, new signals grew, including a triplet at 4.4 ppm and a thin doublet-like signal at ca. 2.9 ppm, both characteristic of Gly-Cys 3a. We repeated this experiment many times, generally at a concentration of 5 × 10 −3 to 5 × 10 −2 mol/L, for practical NMR measurements. However, we also tested it at 3 10 −4 mol/L, a concentration at which 2a and 3a were also obtained.  We have studied the influence of the pH on these reactions. The results are summarized in Figure  3. The ring formation is quicker under basic conditions. We believe that under such conditions, the thiol function is deprotonated, giving the more nucleophilic thiolate species. Under an acidic condition, the nucleophilic species is probably the thiol itself. The hydrolysis step is quicker under acidic conditions. This probably implies that the thiazoline ring is activated through protonation of the double bonded nitrogen atom before H2O addition. We have also tested these reactions at 24 °C and 70 °C. Not surprisingly the process is quicker at a higher temperature, but also goes well at room temperature.
The conditions in the ocean four billion years ago are not precisely known. However, water was probably still hotter than now [27] and the presence of large amounts of CO2 in the atmosphere might imply that it was slightly acidic (nowadays, ocean's pH is 8.1) [28]. Taking these considerations into account, we chose a temperature of 45 °C and a pH of ca. 5.5-6.5 as standard conditions. We have studied the influence of the pH on these reactions. The results are summarized in Figure 3. The ring formation is quicker under basic conditions. We believe that under such conditions, the thiol function is deprotonated, giving the more nucleophilic thiolate species. Under an acidic condition, the nucleophilic species is probably the thiol itself. The hydrolysis step is quicker under acidic conditions. This probably implies that the thiazoline ring is activated through protonation of the double bonded nitrogen atom before H 2 O addition.  We have studied the influence of the pH on these reactions. The results are summarized in Figure  3. The ring formation is quicker under basic conditions. We believe that under such conditions, the thiol function is deprotonated, giving the more nucleophilic thiolate species. Under an acidic condition, the nucleophilic species is probably the thiol itself. The hydrolysis step is quicker under acidic conditions. This probably implies that the thiazoline ring is activated through protonation of the double bonded nitrogen atom before H2O addition. We have also tested these reactions at 24 °C and 70 °C. Not surprisingly the process is quicker at a higher temperature, but also goes well at room temperature.
The conditions in the ocean four billion years ago are not precisely known. However, water was probably still hotter than now [27] and the presence of large amounts of CO2 in the atmosphere might imply that it was slightly acidic (nowadays, ocean's pH is 8.1) [28]. Taking these considerations into account, we chose a temperature of 45 °C and a pH of ca. 5.5-6.5 as standard conditions. We have also tested these reactions at 24 • C and 70 • C. Not surprisingly the process is quicker at a higher temperature, but also goes well at room temperature.
The conditions in the ocean four billion years ago are not precisely known. However, water was probably still hotter than now [27] and the presence of large amounts of CO 2 in the atmosphere might imply that it was slightly acidic (nowadays, ocean's pH is 8.1) [28]. Taking these considerations into account, we chose a temperature of 45 • C and a pH of ca. 5.5-6.5 as standard conditions. Under such conditions, we observed no reaction between aminoacetonitrile and any other proteinogenic amino acid that we tested (glycine, alanine, serine, methionine, aspartic acid, histidine, and lysine). It is worth noting that serine did not react. It appears that its alcohol function is not nucleophilic enough to attack the CN triple bond. Hence, the presence of a thiol function is mandatory. Indeed, homocysteine (Hcy) did react with a reaction rate similar to that observed with cysteine. In this case, the intermediate is the six-membered ring 4a, and the final product is Gly-Hcy 5a (Figure 4). Under such conditions, we observed no reaction between aminoacetonitrile and any other proteinogenic amino acid that we tested (glycine, alanine, serine, methionine, aspartic acid, histidine, and lysine). It is worth noting that serine did not react. It appears that its alcohol function is not nucleophilic enough to attack the CN triple bond. Hence, the presence of a thiol function is mandatory. Indeed, homocysteine (Hcy) did react with a reaction rate similar to that observed with cysteine. In this case, the intermediate is the six-membered ring 4a, and the final product is Gly-Hcy 5a (Figure 4).  Some other representative results are presented in Figure 5. They show that the acid function of cysteine can be replaced by a primary or secondary amide. When Cys-Gly was used, the tripeptide Gly-Cys-Gly 9 was obtained with a very good conversion. N-Acetyl aminoacetonitrile 1b, which can be considered as a model for any other N-acyl acetonitrile, including cyano-terminated peptides, also reacted with a good rate (Figure 6a). In contrast, the reaction was slower when aminoacetonitrile was replaced by β-aminopropionitrile 1c (Figure 7). In these two last examples, the hydrolysis step was quick. No reaction was observed with the γ-nitrile of glutamic acid 1d [29]. The selectivity in favor of α-aminonitriles was also exemplified when the bis-nitrile derived from aspartic acid 1e [30] was used. In this case, only the α-aminonitrile reacted, giving the corresponding thiazoline 2e, which was stable under these conditions (Figure 6c,d).

CN
Finally, we also tested the reactivity of penicilamine, a sterically hindered aminothiol. In this case, the reaction was very slow, probably because of the bulkiness of the gem-dimethyl substituents. Furthermore, the only detected product was the final dipeptide 10 (Figure 6b). This might be due to the electron donating property of the methyl groups, making the nitrogen atom of the intermediate thiazoline ring more basic. Protonation of this nitrogen atom would thus be easier, hence the hydrolysis step quicker.
In order to explain the observed selectivity, we calculated the level of the π* orbital of a series of nitriles ( Table 1).
The lowest calculated orbital was that of the protonated form of aminoacetonitrile. Such a level would explain its greater reactivity compared to other nitriles. Noticeably, the non-protonated form of aminoacetonitrile is predicted to be much less reactive and so is probably not involved in the Some other representative results are presented in Figure 5. They show that the acid function of cysteine can be replaced by a primary or secondary amide. When Cys-Gly was used, the tripeptide Gly-Cys-Gly 9 was obtained with a very good conversion. N-Acetyl aminoacetonitrile 1b, which can be considered as a model for any other N-acyl acetonitrile, including cyano-terminated peptides, also reacted with a good rate (Figure 6a). In contrast, the reaction was slower when aminoacetonitrile was replaced by β-aminopropionitrile 1c (Figure 7). In these two last examples, the hydrolysis step was quick. No reaction was observed with the γ-nitrile of glutamic acid 1d [29]. The selectivity in favor of α-aminonitriles was also exemplified when the bis-nitrile derived from aspartic acid 1e [30] was used. In this case, only the α-aminonitrile reacted, giving the corresponding thiazoline 2e, which was stable under these conditions (Figure 6c,d).
Finally, we also tested the reactivity of penicilamine, a sterically hindered aminothiol. In this case, the reaction was very slow, probably because of the bulkiness of the gem-dimethyl substituents. Furthermore, the only detected product was the final dipeptide 10 (Figure 6b). This might be due to the electron donating property of the methyl groups, making the nitrogen atom of the intermediate thiazoline ring more basic. Protonation of this nitrogen atom would thus be easier, hence the hydrolysis step quicker. In order to explain the observed selectivity, we calculated the level of the π* orbital of a series of nitriles (Table 1).
The lowest calculated orbital was that of the protonated form of aminoacetonitrile. Such a level would explain its greater reactivity compared to other nitriles. Noticeably, the non-protonated form of aminoacetonitrile is predicted to be much less reactive and so is probably not involved in the reaction mechanism. Also, the π* orbital of β-propionitrile is higher (it is less reactive) and the simplest γ-aminonitrile is predicted to be even less reactive (no reaction from glutamic nitrile). In contrast, α-substitution of aminoacetonitrile, as in α-propionitrile, should not significantly alter its reactivity. reaction mechanism. Also, the π* orbital of β-propionitrile is higher (it is less reactive) and the simplest γ-aminonitrile is predicted to be even less reactive (no reaction from glutamic nitrile). In contrast, α-substitution of aminoacetonitrile, as in α-propionitrile, should not significantly alter its reactivity.    We studied this substitution effect using the nitriles derived from two other amino acids ( Figure  8). L-Methionine nitrile 1f [31] was prepared from N-protected L-methionine in a three-step process.   We studied this substitution effect using the nitriles derived from two other amino acids ( Figure  8). L-Methionine nitrile 1f [31] was prepared from N-protected L-methionine in a three-step process.  We studied this substitution effect using the nitriles derived from two other amino acids (Figure 8). L-Methionine nitrile 1f [31] was prepared from N-protected L-methionine in a three-step process. Valine nitrile 1g, was prepared as a racemic mixture using a Strecker reaction from the corresponding aldehyde [32]. Valine nitrile 1g, was prepared as a racemic mixture using a Strecker reaction from the corresponding aldehyde [32]. Their reaction with cysteine amide [33] was studied (Figure 8). In these cases, we were able to isolate the intermediate cycles in pure form (as a 1/1 mixture of diastereoisomers from racemic ValCN). The deceptive isolated yields, despite the slightly basic conditions we used, which should have slowed the hydrolysis step, were probably due to important hydrolysis during column chromatography on silica. In addition, we found that the rate of hydrolysis in water of the valinederived thiazoline 12 was much slower than the one of the methionine derivative 11 (and of the simplest Glycine derivative 2a). This is probably due to the presence of the bulky isopropyl group in 12. On the basis of our experiments, we propose that AA-Cys and AA-Hcy dipeptides were overrepresented in the primitive ocean (compared to non-thiol-containing dipeptides).
These dipeptides are thiols and as such, could be major players in a "thioester world" [34]. Indeed, when we mixed Gly-Cys 3a (obtained from a 1 to 1 mixture of GlyCN and cysteine) with an excess of GlyCN in D2O solution at 45 °C (Figure 9), a peak was observed at 194.16 ppm in the 13 C NMR of the reaction mixture ( Figure 10). Such a chemical shift is characteristic of the thioester function. We believe that it belongs to compound 14. We also noticed the formation of glycine amide 15. These products would both derive from the first formed C=N double-bonded addition product 13. The thioester was partly hydrolyzed to give glycine, but we were also able to characterize, among the reaction products, the amidonitrile Gly-GlyCN 1h ( 13 C NMR: 27.58, 40.34, 116.74, 167.58 ppm), meaning that the thioester reacted with the non-protonated amino group of GlyCN (which is possible because of the low pKa of GlyCN: 5.55 [35]). For instance, in an experiment in which we used globally 4 eq. of GlyCN (relative to cysteine), after two days at 45 °C, the observed GlyOH/GlyNH2/GlyGlyCN ratio was found to be 21/37/42. This demonstrates that Gly-Cys is able to promote the formation of a peptide bond from a nitrile. Similar results were obtained for Gly-Hcy. In addition, as our theoretical calculations predicted (Table 1), when Gly-GlyCN 1h was mixed with cysteine, Gly-Gly-Cys 3h [36] was readily formed ( Figure 5), demonstrating that not only dipeptides, but also tripeptides, could have been formed by this process in the primitive ocean. Their reaction with cysteine amide [33] was studied ( Figure 8). In these cases, we were able to isolate the intermediate cycles in pure form (as a 1/1 mixture of diastereoisomers from racemic ValCN). The deceptive isolated yields, despite the slightly basic conditions we used, which should have slowed the hydrolysis step, were probably due to important hydrolysis during column chromatography on silica. In addition, we found that the rate of hydrolysis in water of the valine-derived thiazoline 12 was much slower than the one of the methionine derivative 11 (and of the simplest Glycine derivative 2a). This is probably due to the presence of the bulky isopropyl group in 12.
On the basis of our experiments, we propose that AA-Cys and AA-Hcy dipeptides were over-represented in the primitive ocean (compared to non-thiol-containing dipeptides).
These dipeptides are thiols and as such, could be major players in a "thioester world" [34]. Indeed, when we mixed Gly-Cys 3a (obtained from a 1 to 1 mixture of GlyCN and cysteine) with an excess of GlyCN in D 2 O solution at 45 • C (Figure 9), a peak was observed at 194.16 ppm in the 13 C NMR of the reaction mixture ( Figure 10). Such a chemical shift is characteristic of the thioester function. We believe that it belongs to compound 14. We also noticed the formation of glycine amide 15. These products would both derive from the first formed C=N double-bonded addition product 13. The thioester was partly hydrolyzed to give glycine, but we were also able to characterize, among the reaction products, the amidonitrile Gly-GlyCN 1h ( 13 C NMR: 27.58, 40.34, 116.74, 167.58 ppm), meaning that the thioester reacted with the non-protonated amino group of GlyCN (which is possible because of the low pKa of GlyCN: 5.55 [35]). For instance, in an experiment in which we used globally 4 eq. of GlyCN (relative to cysteine), after two days at 45 • C, the observed GlyOH/GlyNH 2 /GlyGlyCN ratio was found to be 21/37/42. This demonstrates that Gly-Cys is able to promote the formation of a peptide bond from a nitrile. Similar results were obtained for Gly-Hcy. In addition, as our theoretical calculations predicted (Table 1), when Gly-GlyCN 1h was mixed with cysteine, Gly-Gly-Cys 3h [36] was readily formed ( Figure 5), demonstrating that not only dipeptides, but also tripeptides, could have been formed by this process in the primitive ocean. However, the formation of other products was also evidenced in the reaction of cysteine with excess GlyCN. Thus, in the 13 C NMR spectra, peaks at 130-140 ppm were observed ( Figure 10). The mass spectrum of a reaction mixture in water also showed the formation of various products ( Figure  11, see supplementary materials for complete spectrum and further mass attributions). This mass spectrum first confirmed the presence of Gly -GlyCN 1h (protonated, found 114.0656, calcd 114.0667). Another mass was detected at 113.0817. It could be attributed to the amidine 16 (calcd 113.0827). However, in accordance with the observed 13 C NMR of the mixture, we propose that this amidine cyclized and that this mass peak should be attributed to the imidazole 18. Indeed, the cyclization of an amidinonitrile similar to 16 into an aminoimidazole (5-amino-2-methyl-1H-imidazole) has already been reported [37]. At least some of the other products observed in the mass spectrum would be evolution products of 18. For instance, this imidazole could lose ammonia to give the stabilized cation 19 that would in turn react with 18 (in its free amine form) to yield the bis-imidazole 20 (M+H, found 208.1300, calcd 208.1310). 18 itself could react with thiolester 14 to give the amide 21. Other structures are possible (see Supplementary Materials). However, the formation of other products was also evidenced in the reaction of cysteine with excess GlyCN. Thus, in the 13 C NMR spectra, peaks at 130-140 ppm were observed ( Figure 10). The mass spectrum of a reaction mixture in water also showed the formation of various products (Figure 11, see Supplementary Materials for complete spectrum and further mass attributions). This mass spectrum first confirmed the presence of Gly-GlyCN 1h (protonated, found 114.0656, calcd 114.0667). Another mass was detected at 113.0817. It could be attributed to the amidine 16 (calcd 113.0827). However, in accordance with the observed 13 C NMR of the mixture, we propose that this amidine cyclized and that this mass peak should be attributed to the imidazole 18. Indeed, the cyclization of an amidinonitrile similar to 16 into an aminoimidazole (5-amino-2-methyl-1H-imidazole) has already been reported [37]. At least some of the other products observed in the mass spectrum would be evolution products of 18. For instance, this imidazole could lose ammonia to give the stabilized cation 19 that would in turn react with 18 (in its free amine form) to yield the bis-imidazole 20 (M+H, found 208.1300, calcd 208.1310). 18 itself could react with thiolester 14 to give the amide 21.   In order to further ascertain the cyclization of the imino-compound 16, we calculated its stability in comparison to cyclized forms. We used the strategy described previously for Table 1 results, with the 6-31+G(d,p) basis set ( Figure 12). Not surprisingly, 18 was calculated to have free enthalpy 12.7 kcal/mol lower than 16 and 6.9 kcal/mol more stable than 17.
Interestingly, it was found that the dissociation of 18 into 19 + NH3 only costs around 7 kcal/mol. The ΔH of dissociation is nearly 19 kcal/mol (18.67 kcal/mol), but due to the dissociative character of  In order to further ascertain the cyclization of the imino-compound 16, we calculated its stability in comparison to cyclized forms. We used the strategy described previously for Table 1 results, with the 6-31+G(d,p) basis set ( Figure 12). Not surprisingly, 18 was calculated to have free enthalpy 12.7 kcal/mol lower than 16 and 6.9 kcal/mol more stable than 17.
Interestingly, it was found that the dissociation of 18 into 19 + NH3 only costs around 7 kcal/mol. The ΔH of dissociation is nearly 19 kcal/mol (18.67 kcal/mol), but due to the dissociative character of In order to further ascertain the cyclization of the imino-compound 16, we calculated its stability in comparison to cyclized forms. We used the strategy described previously for Table 1 results, with the 6-31+G(d,p) basis set ( Figure 12). Not surprisingly, 18 was calculated to have free enthalpy 12.7 kcal/mol lower than 16 and 6.9 kcal/mol more stable than 17.
Interestingly, it was found that the dissociation of 18 into 19 + NH 3 only costs around 7 kcal/mol. The ∆H of dissociation is nearly 19 kcal/mol (18.67 kcal/mol), but due to the dissociative character of the process, the ∆G value drops to 7.01 kcal/mol. This process does not show a well-defined TS. Thus, we think that the formation of cation 19 proposed in Figure 9 is a plausible event.
We studied more precisely the cyclization step from 16 to 17 ( Figure 13). H 3 O + was used to promote the reaction and to give a proton to the nitrogen atom of the nitrile group, which becomes part of the exocyclic imine of 17. Two explicit molecules of water were introduced, in addition to the water continuum. It appeared that the cyclization step should be exocyclic and quick, with a low level TS (activation energy of 6.6 kcal/mol). This is another confirmation that the compound of mass 113.0817 we observed was not 16, but indeed the imidazole 18 (resulting from a simple proton migration from 17). It is noticeable that similar calculations for the potential cyclization of the amide 1h into an oxazole indicated that this reaction should not happen. Indeed, we never observed it experimentally. In sharp contrast with 16, 1h (GlyGlyCN) is stable. the process, the ∆G value drops to 7.01 kcal/mol. This process does not show a well-defined TS. Thus, we think that the formation of cation 19 proposed in Figure 9 is a plausible event.
We studied more precisely the cyclization step from 16 to 17 ( Figure 13). H3O + was used to promote the reaction and to give a proton to the nitrogen atom of the nitrile group, which becomes part of the exocyclic imine of 17. Two explicit molecules of water were introduced, in addition to the water continuum. It appeared that the cyclization step should be exocyclic and quick, with a low level TS (activation energy of 6.6 kcal/mol). This is another confirmation that the compound of mass 113.0817 we observed was not 16, but indeed the imidazole 18 (resulting from a simple proton migration from 17). It is noticeable that similar calculations for the potential cyclization of the amide 1h into an oxazole indicated that this reaction should not happen. Indeed, we never observed it experimentally. In sharp contrast with 16, 1h (GlyGlyCN) is stable.   the process, the ∆G value drops to 7.01 kcal/mol. This process does not show a well-defined TS. Thus, we think that the formation of cation 19 proposed in Figure 9 is a plausible event.
We studied more precisely the cyclization step from 16 to 17 ( Figure 13). H3O + was used to promote the reaction and to give a proton to the nitrogen atom of the nitrile group, which becomes part of the exocyclic imine of 17. Two explicit molecules of water were introduced, in addition to the water continuum. It appeared that the cyclization step should be exocyclic and quick, with a low level TS (activation energy of 6.6 kcal/mol). This is another confirmation that the compound of mass 113.0817 we observed was not 16, but indeed the imidazole 18 (resulting from a simple proton migration from 17). It is noticeable that similar calculations for the potential cyclization of the amide 1h into an oxazole indicated that this reaction should not happen. Indeed, we never observed it experimentally. In sharp contrast with 16, 1h (GlyGlyCN) is stable.

Conclusions
A world containing small peptides and heterocycles, based on the chemistry of thiols and nitriles, can be delineated. It could have persisted as long as a significant amount of HCN was present in the ocean and permitted the synthesis of aminonitriles from aldehydes. In this "cyano-sulfidic" world [14], thiol-containing peptides would have been the most important molecules. We propose to name it the "Thiol Rich Peptide (TRP) world" [8]. In such a world, not only dipeptides, but also tripeptides, would have been formed. For instance, any dipeptide nitrile AA1-AA2CN (the simplest example being Gly-GlyCN) produced from the reaction of H 2 NAA2CN with a thioester of AA1, would react with cysteine to give the tripeptide AA1-AA2-Cys, and with homocysteine to give AA1-AA2-Hcy. Could some of these tripeptides have been the very first catalytic triades [38]? Indeed, we have demonstrated that even dipeptides like GlyCys (but not monomeric cysteine) are able to promote the formation of peptide bonds from nitriles. They are also able to promote the formation of imidazoles. Such heterocycles play an important role in today's biochemistry. Of special interest is the simplest aminoimidazole, which, as its ribonucleotide derivative (AIR) [39], is an intermediate in the de-novo synthesis of inosine monophosphate (IMP), hence of purine nucleotides. Thus, imidazoles could have established a bridge from peptides to nucleic acids.