This article is from
Journal of Creation 36(3):72–80, December 2022

Browse our latest digital issue Subscribe

Racemization of amino acids under natural conditions: part 4—racemization always exceeds the rate of peptide elongation in aqueous solution

by Royal Truman and Boris Schmidtgall

Cells cannot function without several hundred kinds of proteins, each in multiple copies. Researchers have spent enormous effort trying to find naturalist scenarios to produce these very large peptides, which result from the condensation of amino acids (AAs). Reading the origin of life (OoL) literature, we have noticed a recurring principle. The experimental conditions designed to facilitate AA condensation would automatically also facilitate racemization.

Here, in part 4 of this series, we will propose a startling hypothesis assuming plausible naturalistic conditions in water.

Hypothesis: On average, peptides derived from pools of pure L-amino acids (L-peptide_n) will convert one or more of their n L-residues (L → D) before elongating by one amino acid. This holds for all peptide sizes and temperatures.

This is a remarkable claim. It means that a family of peptides of any length would be assumed to contain only L-residues initially and would be surrounded by a pool of only L-AAs. If every step in such a chain of events were to decrease enantiopurity faster than peptide elongation, then very large enantiopure proteins would not be produced naturally. Introducing more realistic conditions further cements this conclusion, since each sequential elongation would inherit the D-residues of its predecessor peptide plus add D-enantiomers resulting from conversion of any L-enantiomeric excess. As a racemic state [L] ≈ [D] is approached, further net L → D would be slower due to the D → L back reaction, as discussed in part 2 of this series.¹ This of course does not contradict our hypothesis. Long-term racemization in aqueous solution is unavoidable absent specialized enzymes and highly designed processes such as those used by cells.

Our hypothesis stipulates that the initial peptide consists of only n residues. We are obviously not claiming that at least one more L-enantiomer will always be converted to a D-enantiomer for all peptides. Should the initial peptide_n consist of [L] ≈ [D], this claim would imply that [D] > [L] would result, which everyone will agree is nonsense. Instead, we are arguing that L-peptide_n–3 would result in peptide_n–2 having one or more residues converted to D. If these would be magically reset to L before subsequent elongation, the new pristine L-peptide_n–2 would have one or more residues converted to D during the process of condensation with an AA to form peptide_n–1.

Admittedly, the resetting of all generated D back to L will increase the number of positions where L → D could subsequently occur. Clearly, however, the increased number of L residues cannot compensate for having just conceptually eliminated that same number of already fully converted D residues. Since this would hold for n of all sizes, initiating condensation from a pool of pure proteinogenic AAs cannot produce large peptides having high L content, as required for biological proteins.

We can make this conclusion more plausible in an easily understood manner. Consider only the end position of peptide_n. Once an L-AA is added, that new end residue has a higher probability of being converted to D before the next elongation could occur, since residue inversion is kinetically a first order reaction, whereas elongation is second order, as shown in figure 1.

Furthermore, that end residue has a higher probability of being lost through hydrolysis than subsequent further elongation, and fragments generated from hydrolyzation retain their L or D states. Although this is sufficient to demonstrate the point, a more realistic perspective will help remove any possible doubts. Internal bonds can also racemize but can’t condense to form larger linear peptides. Furthermore, a fully formed carbanion intermediate is not necessary for racemization to occur.

If we are right, this would pose a serious dilemma for the OoL community, since we showed in part 1 that only about 5–10% D-amino acids need be present to preclude secondary structures from forming in putative ancestral peptides.² Folded proteins cannot be produced without secondary structures.

Trust in deep time to offer opportunities for fortuitous chemical reactions in the face of both unfavourable Gibbs free energy and kinetic rates for amino acid condensation is a conceptual mistake. More time would inevitably increase the density of contaminating racemized AAs in both free form state and chemically bound.

In part 2, we documented the available kinetic and thermodynamic data we were able to find for AA racemization and in part 3 the data related to AA condensation.¹

Now we must combine data on condensation and racemization, since only large homochiral peptides can fulfill the necessary life-related chemical processes. Ideally the conditions under which peptide elongation and racemization occur should be the same to facilitate the analysis, in the best case as part of the same experiments. Since unfortunately this is rarely if ever done, interpolation and extrapolations must be made here.

We used the Arrhenius equation and data measured by Steen et al.³ to calculate rate constants for L → D of various AAs at 250°C (see table 1), since data is available for amino acid condensation at this temperature.

The amino acid inversion rate constants ranged from 0.19 to 89 /hour, in sharp contrast with the Steen et al. experiments which, after considerable optimization, produced a mere ~0.001% (gly)₈ condensation product after about two hours. Clearly loss of homochirality at 250°C is many orders of magnitude faster than peptide elongation. We saw in part 2 how small temperature changes greatly accelerate AA racemization.³ Our analysis of all the data we have been able to access indicates that racemization and condensation slow considerably at ~0°C, but increasing temperature increases racemization far more rapidly than condensation. Furthermore, hydrolysis of peptide_n also increases rapidly with temperature, hindering formation of all larger peptides, but not compensating by any increase of excess L-AAs.

Facilitating formation of peptides accelerates racemization

The peptide end residues are important for the purpose of this paper, which is to compare rates of peptide elongation vs L → D inversion. Producing larger peptides requires reacting an amino acid with either end of an existing peptide. Interestingly, the relative rates of racemization for proteinogenic AAs in peptides follow the relation:

NH₂-terminal > diketopiperazine » COOH-terminal ≈ interior positions ≈ free AA.      [1]

Hydrolysis of an end residue produces a stable zwitterion, so these occur more readily than internal peptide bonds.^4,5 Cleavage of an internal peptide bond, however, produces two end residues, which become candidates for hydrolysis. This is another reason that forming large oligomers is so difficult.

That peptide residues racemize faster than free AAs presents a dilemma which naturalists may not be aware of. Ever more ways to facilitate formation of peptides in water might seem useful to create chance opportunities for some life-relevant process to occur, but the more peptides formed, the faster an initial reservoir of excess L-AA would become racemized. The flat carbanion H₃N⁺-C^-R-C is stabilized and therefore can form more easily, permitting L → D inversions. In addition, in-chain racemization is especially fast for some residues such as Asn, Asp⁶ and Ser⁷ through intramolecular racemization catalysis.

I. Peptide growth vs racemization: ΔG is unfavourable

OoL models require large homochiral peptides in high concentrations. But for peptides of any length n, elongation and residue racemization occur simultaneously. Our focus here will be on peptide formation in the absence of created cellular machines such as ribosomes.

Suppose a source of pure L-AAs were available to growing chains of peptides. Condensation of every peptide bond, the reverse of hydrolysis, would be thermodynamically unfavourable, in the case of glycine (Gly) by +2.3 to +3.6 kcal/mole for each peptide bond⁸ or 3.0 kcal/mole according to others (see table 3).^9-11 For those more familiar with SI units, 1 kcal = 4.184 kJoule. Clearly the equilibrium for dissolved peptides favours [peptide_n] « [peptide_n–1].

In contrast, L → D inversion for residues in peptides is rarely unfavourable, with ΔG ≈ 0.

Case 1: random coil peptides

For small peptides in aqueous solution residue, inversion L → D has ΔG = 0, except for some cases of negligible frequency such as presence of stabilizing secondary structure. Random sequence peptides having lengths up to ~20 residues (the largest peptide produced in water using the non-chiral glycine¹²) will almost all form random coils in water. With the relative concentration of [peptide]_n/ [peptide]_n–1 dropping by about a factor of 50 per residue added,⁸ realistically OoL researchers need only focus on Case 1. To illustrate, under realistic abiotic conditions, the molar proportion of [peptide]₂₁ per molar AA would only be ~1.3 × 10^–35 ((1/400)(1/50)¹⁹).¹³ The concentration of larger peptides would continue to decrease rapidly with size.

Changes in temperature or use of catalysts cannot change the fact that hydrolysis of peptides in water is favoured thermodynamically and hinders peptide elongation.¹⁴ Amino acid inversion does not face this barrier. Destruction of peptides could be slowed down by removing water, but for abiogenesis purposes permanent isolation from water would be equivalent to rendering the peptides no longer existent for OoL-relevant chemistry, and racemization could continue, although more slowly.

Conclusion. For small, random-sequence peptides ~20 residues or less which lack a stable secondary structure, the ΔG for residue inversion will always be less than the ΔG for elongation.

Case 2: peptides with secondary structure

For large enough polypeptides derived from random sequences based on pure L-amino acids, occasionally secondary structures such as α-helices and β-sheets could form. For these the ΔG of condensation should be slightly less unfavourable than +2.3 to +3.6 kcal/mole based on glycine oligomerization,⁸ due to a contribution of –ΔG_2s (the free energy due to stabilization by a secondary structure). However, L → D inversions with ΔG ≈ 0 would precede the last AA addition, which led to some secondary structure, seriously hindering these from forming.

Perhaps the entire peptide forms an α-helix or β-sheet. Could an inversion-intolerant L-peptide_n sequence be conceived, possibly already with one or more secondary structures, such that every possible inversion would be less favourable than elongation? This would mean that inversion of every residue, including the end residue where condensation is to occur, must have a ΔG > +2.3 kcal/mole.¹⁵

What would the maximum ΔG_2S penalty be for a residue inversion which disrupts a secondary structure? We could not find the necessary data for a precise answer but can infer that it should rarely, if ever, be more than for peptide elongation, certainly not for every peptide in the theoretical inversion-intolerant peptide. Let us explain why.

The free energy difference between the native folded and denatured states of globular proteins is surprisingly small, typically around –7 kcal/mol at 25°C, with most lying in the range of –5 to –15 kcal/mol.^16,17 Proteins are believed to have optimal sequences to produce stable secondary structures, so peptide sequences able to fold would inevitably have a lower free energy of folding. Assuming ~60% of an average size globular protein (~300 residues) is part of a secondary structure would indicate a ΔG_2S of ~ –0.04 kcal/mol per residue at 25°C (i.e. –7 kcal/mol / 180 residues).

This suggests that a secondary structure involving twenty residues would contribute a ΔG_2S ≈ –0.8 kcal/mole (–0.04 × 20). The penalty for disrupting a secondary structure would therefore be considerably lower than for condensation, and, once disrupted, subsequent inversions would be back to ΔG ≈ 0. Multiple inversions would accumulate, and hydrolysis would dominate with hydrolysis 50 times more probable than elongation.

Conclusion. For larger random-sequence peptides containing secondary structures not deliberately designed to resist residue inversions, the ΔG for residue inversion will be less than the ΔG for elongation.

We suspect it would be possible to design an exceedingly inversion-intolerant peptide_n, which is why we formulated the conclusion as we did. But this would not have arisen naturalistically and would be an example vanishingly unlikely to be useful for sensible OoL purposes.¹⁸

Our scenario presumed an absurd initial environment of a small amount of 100% pure L-AAs with no possibility of contamination for millions of years. We will retain this OoL-friendly scenario but explain in table 2 why a sophisticated inversion-intolerant peptide won’t arise naturalistically.

At what peptide size would one need to consider the presence of secondary structures at all? These structures require enough residues for a measurable amount of stability.

In one experiment Brack and colleagues were able to induce a helix structure based on designed (Leu-Asp-Asp-Leu)_n-Asp peptides between 13 and 25 residues long if the right concentration of Zn²⁺ was added, but the sensitivity to L → D substitution was not reported.⁷ Brack et al. also showed, in the 1970s, that (Leu-Lys)_n β-sheets are very sensitive to the incorporation of about 5% D-isomer and will only form with seven or more of the correct homochiral residues in a row. Even this was only possible under optimized conditions, such as including the right coordinating metals and an aqueous medium with high ionic strength.^7,19-21

These and similar studies use only a small subset of the appropriate AA residues having the correct hydrophilic–hydrophobic (Hi-Ho) pattern, under unrealistic and optimized conditions. Only weak and fleeting secondary structures were formed, the disruption of which would have a negligible +ΔG.

II. Peptide growth vs racemization: kinetic rates are unfavourable

A fundamental question now arises. ΔG considerations address long-term equilibrium state conditions. But could loss of enantiopurity of L-peptides occur faster short-term than increase in peptide size in naturalistic settings? The most favourable abiogenesis scenario involves a pool of 100% pure AAs initially. A given L-peptide_n faces several possible fates, in addition to decomposing chemically (see figure 1). It could condense with an AA with rate constant k₁; invert one or more L-AAs with rate constant k₂; and hydrolyze an end residue with rate constant k₃ (see figure 2).

We will neglect for the moment other possibilities, such as loss of end amino or carboxyl groups, or dehydration reactions.

peptide_n + AA → peptide_n+1, at a rate k₁ × [peptide_n] × [AA]      [2]

peptide_{n (L = l)} → peptide_{n,(L =1 – 1)}, at a rate k₂ × [peptide_n]      [3]

peptide_n + H₂O → peptide_n–1, at a rate k₃’ × [peptide_n] × [H₂O]      [4]

where k₃ = k₃’[H₂O].

Note that [2] is a second order reaction whereas [3] is only a first order reaction. In [3] the number of L-AAs decreased from l to l–1. In [4] the high concentration of water was incorporated into the k₃ rate constant, since it remains essentially constant. Internal bonds could also hydrolyze besides the end residues. This would strengthen our case, since now the fraction of larger peptides has been decreased even more. Note that any AA inversion would persist if condensation or hydrolysis occurs, whereas the rate of elongation has decreased when hydrolysis occurs. This contributes to residue inversion occurring more frequently than peptide elongation.

As shown above, the Gibbs free energy of amino acid condensation is strongly unfavourable, so the long-term equilibration outcome would be mostly zwitterions in water. Therefore, abiogenesis scenarios must demonstrate that shorter term kinetic effects permit large homochiral peptides to arise temporarily. This means that AA condensation and racemization need to be measured and reported under the same conditions. This is rarely, if ever, done, and the ubiquitous experimental use of glycine hides this fact.

We searched the literature for kinetic data on peptide formation in water and AA racemization in peptides, as shown in table 3.

We could not find literature data for rate constants of condensation reactions such as [5] and [6] and must rely on a reasonable proxy for them:

2 zwitterions → AA₂      [5]

AA_n + zwitterion → AA_n+1      [6]

Our reasoning is as follows. Condensation is in equilibrium with peptide hydrolysis (see figure 3), permitting us to calculate the energy of activation, E_a for condensation as ~+26 kcal/mol.

We propose to use values for k₃ as defined in equation [4] for the hydrolysis reaction peptide_n + H₂O → peptide_n–1 as proxies for k₁ in equation [2], since clearly k₃ » k_1. (Recall that k₃ incorporates [H₂O]. Incidentally, water is also involved in stabilizing the transition state of the condensation reaction.²²) Higher temperature hydrolysis experiments have been conducted using glycine,⁸ and the ln(A) and E_a values needed for the Arrhenius equation were reported, which permits us to calculate k₃ at different temperatures. The E_a of condensation will be the same as for hydrolysis plus 2.3 to 3.6 kcal/mol to bring the zwitterion reactants to a neutral AA state in water.

The ~26 kcal/mol E_a for condensation is greater than the average E_a of racemization for AAs per peptide bond (~23 kcal/mol), table 3. But the E_a is not the only factor affecting rate constants. In the Arrhenius equation

k = Ae^–Ea/RT      [7]

the pre-exponential factor A is interpreted as the number of collisions per second occurring which have the proper orientation to react. For the three partners, i.e. AA, peptide and water molecules, to be properly positioned for the transition state geometry in the condensation reaction will be less frequent by chance than the necessary configuration for hydrolysis, which involves only peptide_n and water to be properly positioned. Therefore, we expect the A term to also support our claim that k₁ « k₃.

This brings us to a simple but fundamental insight. For peptide elongation to outpace racemization, equation [8] would be necessary:

k₁[peptide_n][AA] > k₂[peptide_n]      [8]

Since k₃ >> k₁, we can replace k₁ by our proxy k₃, which implies

k₃[peptide_n][AA] > k₂[peptide_n].      [9]

Since [peptide_n] is the same on both sides, and only L-residues are assumed, the fundamental requirement becomes

k₃[AA] > k₂.      [10]

From table 3 we see that the average k₃ values at 25°C and 150°C are smaller than k₂. To make matters worse, any plausible pre-biotic concentration of proteinogenic AAs in water must be [AA] « 1 M. For example, Bada estimated it to be ~10^–10 M in prebiotic oceans.²³ Therefore, [10] cannot be true, and thus [8] even less so. We should recall that [8] already dramatically favoured an OoL perspective by assuming 100% pure L-AAs in a relevant environment and no contamination by D-AAs until some form of life had arisen.

Conclusion. Even excluding contamination by external D-amino acid, L-peptide_n would add D-amino acid content at a rate faster than peptide elongation in aqueous solution for all plausible reaction conditions.

Taking additional facts into consideration further emphasizes this conclusion.

• L-AAs in the isolated ‘feedstock’ would actually racemize over time.
• The effect of hydrolysis of peptide_n to decrease the size of peptides was neglected in our mathematical analysis, although this generates another end-residue which racemizes faster than elongation. All D-AAs generated in the end-residue which hydrolyze would further enrich the feedstock with D-AAs (i.e. even without external contamination).
• Chemical decompositions of peptide_n through loss of amino or carboxyl groups at the end position would discontinue peptide elongation, but the residues would continue to racemize and hydrolyze, enriching the feedstock with D-AAs. Even though we generously assume all L-peptide_n during the elongation interval consist of pure L-residues, the AA they condense with would be increasingly likely to be a D-AA.
• In part 2 we mentioned many ways racemization can be accelerated under natural conditions, such as the presence of bases and various aldehydes.¹

Concluding comments

That racemization will outpace the increase of [peptide_n] → [peptide_n+1] under naturalistic conditions for all values of n and all temperatures holds for all the data we have analyzed. To illustrate, Bada calculated that at pH 7.6 and 0°C in oceans the presence of small quantities of chelating Cu²⁺ lead to a rate constant for racemization of alanine ≈ 1 × 10^–4/year.²² But under these conditions no detectable amount of peptide would be produced. At the other extreme, at 250°C, racemization rate constants are predicted in the timeframe of hours (table 1), but peptides would be instantly hydrolyzed.

Even having assumed an initial pool of pure L-AAs, we have showed that at every stage of L-peptide_n elongation > 1.0 residues on average would convert to D per increase in oligomer by one residue. We know that only 5–10% D-residues randomly distributed in peptides would ‘ruin’ them for life-related chemistry, but for OoL purposes very large homochiral peptides are indispensible.² We have not modelled racemization of peptides over time initiated from pure L-AAs, only shown that these large homochiral peptides could not have formed naturalistically. To introduce more realism, we should not forget that peptides which elongate would inherit the D-residues which had not reverted to L-residues, and hydrolysis of internal peptide bonds would produce smaller fragments, which would also retain their D-residues. Over time the D/L would approach 1 even if an enantiomeric excess of free L-AA surrounded the peptides. Any L-AA just added to a peptide would face a faster rate of inversion than elongation of yet another L-AA.

If our conclusions hold, forming large peptides with very high enantiomer excess of L-residues will not occur under naturalistic conditions. This is true for thermodynamical equilibrium and kinetic reasons. Therefore, condensation and racemization should be measured under the same experimental conditions in all origin of life related experiments. Otherwise, the relevant question of how to produce large homochiral peptides has not been addressed and the failure to do so will not be apparent to most readers. For example, Cronin’s valuable technology,¹² used to perform systematic evaluations of reaction parameters, should be repeated with other pure proteinogenic L-amino acids instead of glycine and the rate of transformation to D-residues measured over time. Glycine is incapable of forming L-peptides and is therefore irrelevant to abiogenesis. Life could not have originated from a collection of poly-glycines.

It is our deep wish to see detailed experiments carried out for a variety of proteinogenic AAs to quantify rate constants, E_a and ΔG for hydrolysis, racemization, degradation, and elongation steps under all relevant parameter settings. Being operational science, this could be a collaboration among those having different views on origins. Attempts to resolve the dilemma we have posed will surely involve the use of non-water polar solvents and special catalysts, which are however dubious proposals for OoL purposes.

Related Media

Information and the origin of life - Did cells write their own software?