Depsipeptides and O-Acyl Isopeptides for Difficult Sequences

A difficult sequence stalls because the chain folds and aggregates on itself, and after about a century of peptide chemistry that remains the main limitation on what is accessible by chemosynthesis — not only for sequences above some critical length, but for relatively short substrates that form highly folded structures, or that form fibrils and aggregates after final deprotection ^[1]. A pseudoproline attacks that by taking the backbone nitrogen into a ring. The depsipeptide approach takes the amide out of the backbone altogether.

Pick a serine or threonine inside the aggregating stretch and move the acyl group of the preceding residue off that residue's α-nitrogen and onto its side-chain hydroxyl. What was an amide is now an ester, and the α-amino group is left free. The chain is the same atoms rearranged — an ester isomer of the target, which the literature calls a depsipeptide or an O-acyl isopeptide ^[1]. The method was introduced on exactly that basis: assemble the O-acyl isopeptide, then run an O→N intramolecular acyl migration to reach the target, with a reported remarkable improvement of the yields ^[2].

The rearrangement is not exotic. N→O intramolecular acyl migration in Ser- or Thr-containing peptides is a long-known side reaction that interconverts ester and amide bonds, and the O-acyl product can be readily converted back to the original N-acyl form under neutral or slightly basic conditions in aqueous buffer ^[3]. The method is that side reaction, run deliberately and in the useful direction. What follows assumes you already know why a chain aggregates on resin and what a pseudoproline does; it covers what the ester buys, when the switch runs, and where the stereochemistry is exposed.

What one ester actually removes

Two changes happen at the substituted position, and only one of them has been measured. The first is definitional: an ester has no N–H, so one backbone hydrogen-bond donor is deleted from the network a beta-sheet is built from, and the residue's amino group becomes a free, ionisable amine rather than part of the backbone. The second is conformational, and it was the observation the whole method grew out of — the presence of an O-acyl instead of an N-acyl residue within the peptide backbone significantly changed the secondary structure of the native peptide ^[4].

The cleanest structural demonstration outside the amyloid work is on a helix-loop-helix domain analogue. Introducing an O-acyl isodipeptide unit at the putative N-cap serine of the C-terminal helix completely abrogated the analogue's propensity for any secondary and tertiary structure, giving a random coil by CD in non-buffered aqueous solution; the helix-loop-helix fold reappeared as soon as spontaneous O→N intramolecular acyl migration restored the native N-cap serine residue ^[5]. Two things to keep straight about that result. It is a solution measurement on a folded peptide domain, not a measurement of how a peptide-resin swells, so it licenses the conclusion that ordered structure is gone in water — not a number for solvation inside a bead. And the effect was total for that construct from a single ester at a single position, which is the scale of change the method trades on.

The solubility gain has a separate origin. Isomerising the amide to the ester liberates the α-amino group, and that ionised amine is what enhances the water solubility of O-acyl products ^[3]. You are not only deleting a donor; you are inserting a permanent positive charge into the middle of a hydrophobic stretch at any pH where the amine is protonated.

On amyloid beta peptide 1-42 (Aβ1-42), one ester was enough. A single insertion of the isopeptide structure into the 42-residue sequence suppressed the unfavourable nature of that difficult sequence, and the deprotected crude was easily purified by HPLC on the strength of its relatively good purity, narrow elution and reasonable water solubility ^[6]. One number is worth quoting with its experiment kept attached to it: a photo-triggered variant of the same isopeptide, carrying a coumarin-derived photocleavable protecting group, had nearly 100-fold higher water solubility than Aβ1-42 and did not self-assemble, the isomerised structure in its backbone having drastically changed the conformation derived from Aβ1-42 ^[7]. That figure belongs to the photocleavable construct, not to the plain isopeptide, and not to isopeptides in general.

The switch, and the pH window it runs in

The nucleophile that closes the migration is the free α-amino group of the esterified Ser or Thr. Below its pK_a it is an ammonium ion and does nothing, which is the entire reason the isopeptide is handleable. It survives acidolytic cleavage, reversed-phase purification in acidic eluent and lyophilisation: the hydrophilic O-acyl isopeptide can be isolated, for example, as a lyophilised TFA salt, and the target peptide is then obtained quantitatively by a final O-to-N intramolecular acyl migration on exposure to neutral conditions ^[8]. The upper end of the working window is stated in the earlier chemistry as neutral or slightly basic conditions in aqueous buffer ^[3].

The rate, where it has been measured, is fast. For the pH-triggered isopeptide of Aβ1-42, ester placed between Gly25 and Ser26, the peptide adopts and retains a monomeric, random-coil state under acidic conditions and, on change to neutral pH, converts to Aβ1-42 promptly and quantitatively with a half-life of approximately 10 s ^[9]. In the synthetic report on the same target, migration at pH 7.4 was very rapid, no other by-product formation was observed, and the isopeptide was stable under storage conditions ^[6]. Both observations are for that peptide at that site. The residue donating the acyl group and the local sequence are part of the system, so treat the 10 s figure as an existence proof that the shift can be fast, not as a rate constant for your own construct.

The practical consequence is that the last step of the synthesis is a chemical reaction you schedule, not a workup. Acid has to come off before anything happens, and the moment you neutralise is the moment the target exists. What the switch does not do is change the peptide you end up with. The isopeptide's solubility and its refusal to self-assemble are properties of the isomer; once migration restores the native backbone, native behaviour returns on its own timescale. In the same system, the peptide produced in situ was monomeric to begin with, and its oligomerisation, amyloid fibril formation and conformational changes could then be observed over time ^[9]. Concentration, buffer and timing at neutralisation are therefore experimental variables, not incidental handling.

Building the ester in during assembly

There are two ways to get the ester into the chain, and they distribute the risk very differently.

The first is on-resin O-acylation: stop at the resin-bound Ser or Thr, unmask the side-chain hydroxyl and acylate it with the next Fmoc-amino acid. The systematic study of depsipeptide methodology investigated precisely the things that go wrong there — the stability of the depsi units during assembly and workup procedures, the completeness of the O-acylation step, the occurrence of epimerisation of the amino acid activated during O-acylation, and the nature of the side products formed ^[10]. The step is awkward enough that the field engineered it away: isodipeptide units have enabled routine use of the O-acyl isopeptide method by avoiding the often difficult esterification reaction on resin ^[4].

The second route is that pre-formed building block. Forty O-acyl isodipeptide units of the form Boc-Ser/Thr(Fmoc-Xaa)-OH were effectively synthesised in two steps without epimerisation and offered as the building blocks that make routine use of the method possible ^[11]. The naming carries the protection scheme: Boc sits on the Ser or Thr α-amine and stays there through repetitive piperidine treatment until acidolytic cleavage, while the Fmoc on the incoming residue is the handle from which chain elongation continues. On the same principle, the depsipeptide method was modified to allow completely automated syntheses of long-chain depsipeptides via depsidipeptide units, without the need for any interruption by manual esterification procedures ^[10].

Three specifics from the reports are worth carrying to the bench.

• Resin. The protected 26-O-acyl isoAβ1-42 peptide resin was assembled on 2-chlorotrityl chloride resin, which gave minimum side reactions in comparison with other resins ^[6].
• Solvent at the unit coupling. A side reaction that deleted Ser26 from the isopeptide structure was traced to the activation step of the isodipeptide unit and was solvent-dependent: in DMF or NMP an intramolecular side reaction originating from the activated species of the unit occurred, whereas in CHCl₃ or CH₂Cl₂ it was less likely to occur; coupling the unit in CH₂Cl₂ gave the target isopeptide with almost no major side reaction ^[12].
• Diketopiperazine. Applying the depsipeptide method to extremely difficult sequences uncovered a serious side reaction involving diketopiperazine formation, subsequently avoided by the appropriate use of the Bsmoc protecting group ^[10].

Generating the ester later, and getting one you did not order

The same rearrangement runs in the N→O direction, and it does so without being asked ^[3]. During Fmoc-based synthesis of phosphopeptides, isomerisation at threonine was observed to furnish depsipeptides; the rate of N→O acyl migration depended on the sequence context, and depsipeptides formed most rapidly when the phosphotyrosine was located at the +2 position ^[13]. That is a phosphopeptide result measured at threonine, so the +2 rule does not transfer to unphosphorylated sequences. The transferable part is the existence claim: a Ser- or Thr-containing peptide can isomerise during synthesis without being designed to, so a depsipeptide may already be in your crude whether or not you put one there.

Because the two forms are constitutional isomers ^[1], they are isobaric, and mass alone will not tell them apart; they differ chromatographically, and they differ in what they do when the sample is neutralised. That is the practical tell — an unplanned depsipeptide converts to the amide isomer as soon as the eluent stops being acidic, so a peak that moves between injections after a buffer change is a candidate. Deliberate installation during assembly is the only version of this you can place, count and time; post-synthetic isomerisation is a side reaction you detect rather than a route you drive.

Racemisation at the acyl donor

The bond you are forming is an ester between an activated N^α-urethane-protected amino acid and a beta-hydroxyl. A hydroxyl is a weaker nucleophile than an amine, so the step tends to be pushed with stronger activation, a nucleophilic catalyst or longer contact time — the conditions under which an activated, urethane-protected amino acid loses configuration.

How much configuration is lost depends on the activation, and the dependence is steep. Esterifying Fmoc-glycosylated serine and cysteine derivatives onto a hydroxymethyl (Wang) resin, the classical methods of ester bond formation — symmetrical anhydride, TBTU/DMAP with or without HOBt, and pentafluorophenyl esters — gave high percentages of racemisation of the glycosylated serine or cysteine residues, while MSNT gave a high yield of resin substitution together with the lowest percentage of racemisation; the glycosylated cysteine derivative always gave lower racemisation than the serine analogue ^[14]. Scope matters here: those are glycosylated building blocks going onto a resin hydroxyl, not an amino acid going onto a Ser side chain in a growing chain. The geometry differs. The activated species, the catalyst and the base do not.

This is the strongest argument for the pre-formed unit. With Boc-Ser/Thr(Fmoc-Xaa)-OH the ester is made once, in solution, under conditions chosen for that single purpose — the forty-unit set was prepared in two steps without epimerisation ^[11]. The on-resin route moves the same bond formation into your own hands and repeats it at every insertion site, which is why the depsipeptide methodology study listed epimerisation of the amino acid activated during O-acylation among the aspects it had to investigate ^[10]. Note also which junction the amyloid work chose: the ester sits between Gly25 and Ser26 ^[9]. Glycine has no α-stereocentre, so at that particular position there is nothing to epimerise — a convenience of that target, not a property of the method.

Segment condensation inverts the picture. Coupling of depsipeptides via base-assisted activation using HBTU occurred not only without loss of configuration but much faster than the coupling of the corresponding all-amide segments, attributed to the low tendency of the activated depsipeptide units, in contrast to the corresponding peptide segments, to form optically labile oxazolones ^[15]. The cost surfaces elsewhere: when the coupling of long depsipeptide segments proceeded slowly, beta-elimination at the activated depsipeptide unit appeared, to an extent dependent on the presence of base in the system and on the type of solvent. It was completely suppressed by using carbodiimide/HOBt activation in a non-polar solvent (DCM), and in more polar media it was limited by substituting TMP for DIEA during HBTU activation, or by using particular solvent mixtures such as DMSO/toluene ^[15]. So the depsi unit shields the stereocentre and exposes the beta-carbon instead; you trade one failure mode for another and control the new one with solvent and base.

Where this sits next to pseudoprolines

Head to head on synthesis efficiency, the depsipeptide technique was shown to be comparable to that of the well-known pseudoproline technique ^[10]. They are not interchangeable in practice, and the difference is about when the disruption stops.

A pseudoproline ring opens at acid cleavage. Whatever it did for solvation and coupling on resin, it has finished doing by the time the crude reaches the column. The isopeptide ester survives cleavage and is carried through purification as the lyophilised TFA salt, reverting only when you decide to neutralise ^[8]. If the bottleneck is the coupling steps, either tool addresses it. If the bottleneck is that the deprotected crude will not dissolve or will not chromatograph, only the isopeptide is still working at that stage.

Neither is a universal fix and they are not exclusive. The depsipeptide method is applicable to sequences containing residues of serine or threonine ^[1]; with no hydroxy amino acid in the aggregating stretch there is no ester site, and you are back to backbone amide protection or a different strategy entirely. On a genuinely hard target both may be needed at once: a published protocol shows exemplary application of depsipeptide and pseudoproline units for an extremely difficult sequence, the Asn15 analogue of the WW domain FBP28, which is impossible to obtain using the standard protocol ^[16].

What an isopeptide result licenses

A clean chromatogram of the isopeptide is a statement about the isopeptide. Before it becomes a statement about the target you need evidence that migration went to completion. The reported conversions were quantitative under the stated conditions ^{[8, 9]}, but that is a result for those constructs at those sites, not a guarantee for yours; reinject after neutralisation and account for the whole peak area rather than assuming the shift is silent and total.

The reach of the method deserves the hedge its authors gave it. The conclusion drawn from the 42-residue work was that the strategy is applicable in the synthesis of large difficult sequence-containing peptides at least up to 50 amino acids ^[6] — a stated expectation extrapolated from that target class, not a measured ceiling and not a claim about proteins.

Finally, be precise about what you report. The isopeptide's solubility, its random-coil CD signature and its refusal to self-assemble ^{[5, 7, 9]} describe a molecule you deliberately built to be different from the one you want. None of it is evidence about the native peptide's behaviour. The instant the acyl group moves back to nitrogen you own the native sequence again, along with everything that made it a difficult sequence in the first place.