Introduction

In the last years protein amyloids have been the subject of numerous lines of research that have uncovered novel biological functions in all forms of life (Chiti and Dobson 2006; Greenwald and Riek 2010). The formation of amyloid fibrils has also long been associated with many diseases including Alzheimer’s and Parkinson’s diseases (Chiti and Dobson 2006; Greenwald and Riek 2010). Amyloids could also have played an important role in the origin of life, as we and others have recently hypothesized (Carny and Gazit 2005; Dale 2006; Maury 2009; Greenwald and Riek 2012). They can be stable in harsh conditions and have inherent templating and stereoselective properties (Meersman and Dobson 2006; Childers et al. 2009; Omosun et al. 2017). As shown already in many studies, amyloids formed by short peptides can have catalytic activities (Reches and Gazit 2003; Mehta et al. 2008; Lara et al. 2013; Rufo et al. 2014; Zhang et al. 2014; Friedmann et al. 2015; Makhlynets et al. 2016) and are able to form highly organized tube-like structures both in isolation (Omosun et al. 2017) and cooperatively upon interaction with other prebiotically relevant entities like fatty acids (Reches and Gazit 2003; Mehta et al. 2008; Lara et al. 2013; Sánchez-Ferrer et al. 2018; Bomba et al. 2018). Our recent findings, showing that the amyloid structure can support the template-directed synthesis of peptides (Rout et al. 2018), is just one of the latest results that strengthens the hypothesis that amyloids were involved in the origin of life. It should be noted that unraveling the mystery of the origin of life is not a question of a few key experiments but rather the synthesis of many smaller steps into hypotheses that can be further refined and tested.

We previously demonstrated that by forming an ordered aggregate, a complementary pair of amyloidogenic peptides can provide a structural scaffold that enhances peptide elongation via sequence-selective, regio-selective and stereoselective addition of activated amino acids (Rout et al. 2018). These results were obtained using amino acids that had been activated by CDI. CDI provides good yields of peptide addition products with few side reactions and was therefore chosen for the initial experiments. However, in pursuit of our goal of a prebiotic chemical replication, we have now chosen to use carbonyl sulfide (COS), a volcanic gas that has been shown to act as a condensation agent for amino acids (Leman et al. 2004), in the amyloid templating system. The proposed mechanism of amino acid activation via COS has several intermediates, but the most reactive species, the N-carboxy anhydride, is likely to be the same as for the activation by CDI (Leman et al. 2004; Ehler and Orgel 1976). Indeed, our new results indicate that the amyloid template-assisted peptide syntheses with either COS- or CDI-activated amino acids are very similar.

Results and Discussion

Stereoselective Addition of Amino Acids to R(FR)3 Templated by (FE)4

In this study we investigate a system of complementary substrate-template amyloid peptides that was previously shown to have a templating capability (Rout et al. 2018). The substrate peptide RFRFRFR-NH2, abbreviated R(FR)3, is the charge complement of the template peptide Ac-FEFEFEFE-NH2, abbreviated (FE)4. Both peptides have C-terminal amides and the template peptide is acetylated on its N-terminus. Since the substrate has one Phe residue less, the amyloid formed by these two peptides is selective for additions at the N-terminus of the substrate peptide. Both peptides were shown to be soluble above 1 mM at neutral pH, while the mixture of 100 μM of each peptide in 20 mM phosphate buffer at pH 7.4 yields an insoluble amyloid (Rout et al. 2018). In order to assay the stereoselectivity of the R(FR)3/(FE)4 amyloid, racemic Phe was activated with COS and added at concentrations from 50 to 500 μM (25–250 μM of each enantiomer) to either the amyloid aggregate R(FR)3/(FE)4 or the soluble substrate R(FR)3 alone. All reactions contained 1 mM potassium hexacyanoferrate as an oxidizing agent that increases the rate of the condensation. After 30 min, the reactions were solubilized with guanidine hydrochloride (GuHCl) and analyzed by reversed-phase HPLC. In order to properly compare the two activation chemistries, we also performed all of the reactions with the CDI activation as previously reported (Rout et al. 2018). As depicted in the chromatograms in Fig. 1 and tabulated in Table 1, the enhanced stereoselectivity and yield of addition of L-Phe to the substrate is independent of the type of activation (COS or CDI). Also, both chemistries give a significantly enhanced yield of the product (FR)4 formed in the amyloid-templated reactions compared to reactions with soluble substrate. While the stereoselectivity of the templated reactions was similar to the previous results with CDI activation, we were surprised to see that the stereoselectivity of the non-templated soluble peptide was substantially higher than previously reported (Rout et al. 2018), although still significantly lower than in the templated addition reaction (Fig. 1). Upon further investigation we found that the reason for the apparent change in stereoselectivity in the non-templated reaction was a variable loss of the L-Phe addition products, attributed to poor product recovery. Compared to the D-Phe addition product fR(FR)3, the L-Phe addition product (FR)4 has higher binding affinity to the plastic walls of the reaction tubes. This causes a selective partial loss of one of the diastereomeric products which results in an apparent lower diastereomeric excess of the L-Phe addition product. By performing all subsequent reactions in Protein LoBind Tubes (Eppendorf) we were able to get a consistently higher recovery of (FR)4 from all of the reactions, thereby minimizing the effect from the loss of the peptide to the tube walls. The total peptide recovery of each reaction was routinely monitored to control for the potential artifacts of poor product recovery. The total peptide recovery is the amount of substrate peptide and its addition products that were detected by HPLC in the reaction compared to the amount of substrate detected in a control reaction without activation. The main result is the same when using the LoBind tubes: the addition of racemic Phe to the R(FR)3/(FE)4 amyloid is significantly more stereoselective for L-Phe addition than the non-templated reaction with soluble peptide. The enhancement in stereoselectivity in the amyloid-templated reactions gives an approximately 1.5 to 3x increase in the diastereomeric ratio (dr) of L to D-addition products compared to non-templated reaction, regardless of the activation chemistry (CDI or COS) (Fig. 1). As previously observed for CDI activated amino acids, the dr is dependent on the Phe concentration in the reaction, indicating that the reaction with COS follows the same or very similar mechanism as for CDI. We previously reported that the templated reaction is not simply first order with respect to L-Phe, while the non-templated and D-Phe addition reactions are first order with respect to Phe. Based on a kinetic analysis, we have proposed that for the L-Phe addition to the amyloid, the stereoselection involves an additional binding event via a reaction intermediate (Rout et al. 2018).

Fig. 1
figure 1

Stereoselective addition of phenylalanine to the R(FR)3/(FE)4amyloid. HPLC chromatograms of the reactions between DL-Phe and the substrate peptide R(FR)3 in the absence (a and b) and the presence (c and d) of the template peptide (FE)4. The Phe was activated with either COS (a and c) or CDI (b and d). The color of the traces indicate four different concentrations of activated DL-Phe: 50 μM (black); 100 μM (red); 200 μM (blue) and 500 μM (green). The identity of the peaks that have been confirmed by standards and or mass spectrometry are labelled using lower case letters to indicate D-amino acid addition products. The dashed lines connect the peaks of fR(FR)3 and (FR)4 from the same chromatogram, highlighting the enhanced stereoselectivity of the amyloid-templated addition reactions. Peaks indicated by “*” come from the side products formed in COS reactions (described in the main text)

Full size image
Table 1 Stereoselectivity and yield of amino acid addition to R(FR)3 and R(FR)3/(FE)4
Full size table

The absolute yield of the reactions with COS are lower than those with CDI (Table 1) which is to be expected considering the generally lower yields of peptide bonds via COS versus CDI activation. In fact, the templated reaction with CDI-activated Phe is so efficient that compared to the reaction with 200 μM Phe, the 500 μM reaction has a decrease in yield of the single addition product that results from an increased amount of multiple addition products (detected as later eluting peaks in the gradient) (Fig. 2). The lower yields with COS activation can in part be explained by an incomplete activation of the amino acid and larger amount of side products formed when using COS (Fig. 1). 1H NMR detection of the phenylalanine thiocarbamate indicate that approximately only 50% of the amino acid is activated. The two most significant side products are indicated with “*” and “**” in Fig. 1. We found that the “*” side product results from a reaction between only the substrate peptide and K3[Fe(CN)6] and does not involve COS. Also, the amyloid protects the substrate from this reaction and “*” is not detected in the templated reactions. The “**” side product is formed both in reactions with the soluble substrate as well as with the amyloid co-aggregate. Therefore, it was further investigated with mass spectrometry and found to have a mass difference of +C, +O, -2H compared to R(FR)3. We therefore conclude that this side product is likely to be the N-terminal hydantoin, a product that has been previously observed with peptides in COS (Leman et al. 2004).

Fig. 2
figure 2

Multiple addition products in the CDI-activated DL-Phe reactions. HPLC chromatograms of templated (red) and non-templated (black) reactions with 500 μM Phe. The peaks whose identities have been confirmed (Rout et al. 2018) are labeled. The region of the chromatogram with multiple additions is indicated. The initial stereoselection that occurs upon the first Phe addition is reflected in the enhanced dr of the double and triple additions as well

Full size image

We further characterized the stereoselectivity of the amyloid for four other hydrophobic amino acids (i.e. Val, Leu, Tyr and Trp) via both COS and CDI activation. Consistent with the results for Phe reactions, the amyloid displays higher stereoselectivity towards L-amino acid addition than does the soluble peptide (Figs. 3, 4, 5 and 6, Table 1). The highest enhancement of diastereomeric ratio was observed for Phe and Val. For example, with 200 μM Phe the dr of the single addition products increased from 1.8 to 3.8 (COS activation) and from 1.9 to 5.2 (CDI activation). Similarly, with 200 μM Val, the template provided an increase in dr from 1.9 to 6.9 (COS) and from 1.8 to 5.7 (CDI). The smallest increase in stereoselectivity is observed with Trp, for which the soluble substrate has an inherently high L-selectivity: for example, in the CDI activation of 200 μM Trp, the template leads to an increase in dr from 4.3 to 6.6. Intriguingly, in some instances the activation chemistry (COS versus CDI) appears to impact the degree of stereoselectivity in amino acid addition reactions. In particular, both the templated and non-templated additions of Trp are sensitive to the type of activation (Fig. 5, Table 1). The soluble substrate R(FR)3 displays a lower stereoselectivity for COS-activated Trp (dr of 1.2–1.7) than for CDI-activated Trp (dr of 2.9–6.6). These observed differences between COS and CDI could arise from the different reaction pathways and intermediates involved in both systems. Despite their similarities, the two activation pathways are known to lead to very different outcomes for some amino acids. In particular the beta-hydroxy amino acids (Ser and Thr) fail to form peptides at all with CDI activation, instead cyclizing into an unreactive oxazolidine (Ehler et al. 1977). Although the addition reaction with Tyr activated with COS displayed enhanced stereoselectivity in the presence of the template, the low yield rendered the calculation of a dr unreliable for these reactions (Fig. 6, Table 1). The low solubility of Trp and Tyr meant that a smaller dilution of the COS activated amino acid was used in the reactions, thus requiring a larger amount of K3[Fe(CN)6] relative to amino acid (see Methods) and reducing the overall efficiency of the activation.

Fig. 3
figure 3

Stereoselective addition of Val to R(FR)3/(FE)4. HPLC chromatograms of the reactions between DL-Val and the substrate peptide R(FR)3 in the absence (a and b) and the presence (c and d) of the template peptide (FE)4. The Val was activated with either COS (a and c) or CDI (b and d). The colors and symbols are as in Fig. 1

Full size image
Fig. 4
figure 4

Stereoselective addition of Leu to R(FR)3/(FE)4. HPLC chromatograms of the reactions between DL-Leu and the substrate peptide R(FR)3 in the absence (a and b) and the presence (c and d) of the template peptide (FE)4. The Leu was activated with either COS (a and c) or CDI (b and d). The colors and symbols are as in Fig. 1

Full size image
Fig. 5
figure 5

Stereoselective addition of Trp to R(FR)3/(FE)4. HPLC chromatograms of the reactions between DL-Trp and the substrate peptide R(FR)3 in the absence (a and b) and the presence (c and d) of the template peptide (FE)4. The Trp was activated with either COS (a and c) or CDI (b and d). The colors and symbols are as in Fig. 1

Full size image
Fig. 6
figure 6

Stereoselective addition of Tyr to R(FR)3/(FE)4. HPLC chromatograms of the reactions between DL-Tyr and the substrate peptide R(FR)3 in the absence (a and b) and the presence (c and d) of the template peptide (FE)4. The Tyr was activated with either COS (a and c) or CDI (b and d). The colors and symbols are as in Fig. 1

Full size image

Sequence-Selective Addition of Amino Acids to the R(FR)3/(FE)4 Amyloid

To investigate the amino acid selectivity of the R(FR)3/(FE)4 amyloid we set up addition reactions with a mixture of four L-amino acids (Phe, Val, Arg and either Asp or Gly), once again activated either with CDI or with COS. The reactions with Asp and Gly in the mixtures were tested separately because the DR(FR)3 and GR(FR)3 products were not resolved by the HPLC analysis. The reaction mixtures contained 100 μM of peptide and 100 μM of each amino acid. The presence of the template peptide increased the yield of Phe, Val and Arg addition while for Asp it decreased, likely due to loss of favorable electrostatic interactions between Asp and the Arg residues of R(FR)3 that become involved in salt-bridges with (FE)4 (Fig. 7, chromatograms). Gly addition yield was not strongly affected by the presence of the template. A comparison of the relative selectivity for each of the amino acids in the reaction mixtures shows that the presence of the template consistently increased the selectivity of the reaction only for Phe (Fig. 7, bar graphs). In one reaction (Phe, Val, Arg, Asp with CDI) the Val selectivity also increased with addition of the template, but this is primarily the effect of the extreme anti-selection of Asp addition: the non-templated Asp addition was so efficient that it depleted the substrate resulting in a very low relative selectivity for non-templated Val addition (Fig. 7).

Fig. 7
figure 7

Sequence selectivity of amyloid-templated additions. HPLC chromatograms and relative yields from the reaction of R(FR)3 in the absence (black) and presence (red) of the template peptide (FE)4 with a mixture of four amino acids (F, V, R, D in panels a and b or F, V, R, G in panels c and d as indicated). The amino acid mixture was activated with either COS (a and c) or CDI (b and d). Each amino acid was used at 100 μM and its single addition product to the substrate peptide is indicated above its corresponding peak. The red letters in the peptide sequences indicate amino acid residues that were added via the addition reactions. The sequence selectivity of each reaction is represented as the yield of each single addition product relative to the other single addition products

Full size image

Conclusions

Our results touch on two important questions in prebiotic chemistry: the origin of homochirality and the origin of replicating systems. While not a replicating system, the peptides described here provide the framework for a system of complementary cross-templating peptides that together could carry out a replication analogous to the semi-conservative replication of DNA. Mechanisms that break the mirror symmetry to create small enantiomer imbalances are well known and some have even been shown to amplify these imbalances to near homochirality (reviewed in Hein et al. 2013). However, it is not always clear how such mechanisms would apply to a prebiotic scenario. Peptides display significant stereoselectivity for the addition of amino acid NCAs (Blair and Bonner 1980; Hitz and Luisi 2004; Tsuruta et al. 1967) a phenomenon that we also observed in the soluble peptide reactions in this work. The enhancement of this selectivity in β-structured aggregates has also been previously observed and provides a mechanism for the amplification of chirality (Rubinstein et al. 2007; Hitz and Luisi 2004).

In a bottom-up approach to understanding how life could have originated, it is important to find chemically plausible systems that can support metabolic processes and replication. This work expands on our knowledge of prebiotically relevant amyloid systems that provide a mechanism for enhanced selectivity and yield in peptide elongation reactions. The fact that the observed templating mechanism functions similarly despite the differences in activation chemistry indicates that it is robust and could work for a wide variety of peptide bond-forming chemistries. Also, by showing that COS and CDI activation are to a first approximation interchangeable, this finding enhances the prebiotic bona fides for other experiments in which CDI-based activation was used. Under the conditions in this study, the yields of the addition products are lower and more side-products are formed with the more prebiotically relevant COS activation. However, some of the side products are inhibited by the presence of the template, indicating that the amyloid acts to protect the substrate from reactions that would interfere with peptide elongation. Thus, in addition to providing a mechanism for the templated elongation of a peptide, the amyloid also provides a way to stabilize the substrate and protect it from the harsh prebiotic environment, making amyloids an attractive target for further study in the chemical origins of life.

Methods

Reagent Preparation and Purity

The peptides (GLS China) were synthesized using standard Fmoc chemistry on a Rink amide resin. The crude peptides were purified by reverse-phase HPLC on a Kinetex C18 5 μm 10 × 250 mm column (Phenomenex) in CH3CN/H2O/TFA solvent systems, quantitated by their calculated extinction coefficient at 214 nm (Kuipers and Gruppen 2007) and stored in lyophilized aliquots. Before use, the peptides were dissolved in water. Amino acids were purchased with a purity of ≥98% and stock solutions were made in pure water except for Asp, which was adjusted to pH 7.0 with NaOH. Guanidine HCl (8 M, high purity, Pierce) was further purified by passing it over a C18 solid phase extraction column (Supelco) in order to remove a contaminant that eluted close to some of the peptide peaks.

Amino acids were used as purchased, without further purification. The ratios of enantiomers in the racemic amino acids were measured by derivatizing the amino acids with Marfey’s reagent (Bhushan and Brückner 2004) and subsequent reverse-phase HPLC analysis. Within the error of the measurement, Val, Leu and Tyr are racemic while Trp and Phe appeared to have a minor L-excess (3.2% and 3.7% respectively). We further analyzed the DL-Phe by CD spectroscopy. The signal intensity at 225 nm of 7.5 mM DL-Phe was less than 1% of the intensity of 7.5 mM L- or D-Phe. This indicates that the DL-Phe has less than 1% e.e (data not shown).

Amyloid Formation

Lyophilized aliquots of pure peptides were solubilized in water. Peptide co-aggregates were made simply by mixing substrate peptide (100 μM) and a small excess of template peptide (110 μM) and incubating them in Protein LoBind Tube (Eppendorf) at room temperature. Fibrillization was monitored by centrifugation and circular dichroism spectroscopy and appeared to reach equilibrium within a few hours (Rout et al. 2018). The substrate-template peptide mixture formed flocculent aggregates and was briefly sonicated (10 s, 20% power, Bandelin Sonoplus HD 2070 with MS73 microtip) in order to improve liquid handling before setting up the addition reactions.

Amino-Acid Addition Reactions and Analyses

In the CDI reactions, amino acids were activated with two equivalents of carbonyldiimidazole (CDI, ≥ 97% Aldrich) on ice. The individual amino acids were activated at 50 mM except for Tyr (1.2 mM) and Trp (10 mM) due to their low solubility. The mixtures were activated with 40 mM amino acid (10 mM of each amino acid). After 2 min on ice, the activated amino acid solutions were diluted into the peptide mixtures in 20 mM NaPO4 pH 7.4 at room temperature.

For the COS activation reactions, amino acids were dissolved in 200 mM sodium borate solution at pH 9. The amino acid concentrations were the same as for the CDI activations. To an evacuated (15 mbar), sealed flask 1 ml of the amino acid solution and 400–600 mg COS gas (final pressure 1 bar) was added. The solution was briefly mixed by hand several times during activation. After 1 h, the activated amino acid solution was diluted into the 20 mM NaPO4 pH 7.4 buffered peptide mixtures containing 1 mM K3[Fe(CN)6] at room temperature, with the exception of Tyr, for which 4 mM K3[Fe(CN)6] was used. The higher oxidizing agent concentration was needed because the low solubility of Tyr resulted in a higher amount of thiocarbonate in the reactions which also consumes the K3[Fe(CN)6].

After at least 18 h for CDI and after 30 min for COS activated reactions, the reactions were diluted with 3 volumes 8 M guanidine HCl in order to solubilize the peptides for reverse-phase analysis. Reverse-phase analyses were performed on 2.6 μm 4.6 × 150 mm bioZen C18 column (Phenomenex) connected to an Agilent 1200 HPLC system equipped with an auto sampler and diode array detector. The soluble reaction products were injected onto the column and resolved using a linear gradient of acetonitrile with 0.07% TFA (10–40%) at a flow rate of 1.5 ml/min.