C-Terminal Peptide Aldehydes and Ketones: Masked Synthetic Routes

A peptide that ends in an aldehyde cannot be assembled the way a peptide that ends in an amide is assembled. The carbonyl is electrophilic enough to condense with the amines that fill a solid-phase vessel, the carbon beside it epimerises during assembly, and once the compound reaches water it stops being a single species. Every practical route therefore carries a masked precursor through chain elongation and unmasks it at or near the end - sometimes in the cleavage step itself, sometimes by an on-resin oxidation that precedes it. What follows is how those routes differ, what the finished compound looks like on an HPLC trace and in a mass spectrum, and the two applications that make the trouble worth taking.

Why the carbonyl cannot ride along the chain

Take the electrophilicity first. Konno and co-workers acetylated the N-terminus of their target sequences because the free α-amine and the C-terminal aldehyde combine with each other: N-terminal protection, they report, was necessary to avoid (hemi-)aminal formation between the N-terminal amine and the C-terminal aldehyde ^[1]. That is not an exotic side reaction. It is the same condensation solid-phase chemists run on purpose when they load an amine onto an aldehyde-functionalised support: masked amino aldehyde building blocks are attached to backbone amide linker (BAL) resin by reductive amination with sodium cyanoborohydride, and the resulting secondary amine is then acylated with a HATU-activated residue ^[2]. An unmasked aldehyde carried through Fmoc chemistry is an electrophile in a vessel repeatedly flooded with amine base and activated amino acids.

Second, the stereocentre. Kappel and Barany, working with the BAL route, found that the C-terminal residue epimerised by 10% to 25%, which they attributed to the intrinsic stereolability conferred by the aldehyde group rather than to any drawback of the synthesis procedure ^[2]. Guillaumie and colleagues put an independent number on it by making the diastereomeric reference: repeating their route from N^α-Z-D-Phe-OH gave H-Glu-Val-Val-D-Phe-H, and the two free peptide aldehydes resolved cleanly by HPLC. The L-Phe-H peptide contained 12% of the D diastereomer, and the D-Phe-H peptide contained 26% of the L. Importantly, no further racemisation of the L-Phe-H peptide was seen on standing overnight in TFA ^[3]. Where in that process the damage is done was left open: the authors reported that experiments were under way to determine at which stage their compounds racemise. What the overnight experiment rules out is the finished aldehyde standing in TFA. A systematic ¹H NMR study of a model dipeptide aldehyde reached the same place from the other direction: various purification methods were investigated, but none was free from epimerisation ^[4].

Third, the group does not stay an aldehyde. N-acetyl-L-phenylalaninal exists predominantly in its hydrated form in aqueous solution ^[5]. Add an alcohol and you get hemiacetals as well. That is a property of the finished compound rather than the synthesis, taken up separately below — but it is why a masked precursor is easier to handle: an acetal or oxazolidine is one compound with one retention time.

Masked precursors that survive the chain

Acetals and dioxolanes

The most direct approach is to build the C-terminal residue as its acetal and carry that through. Kappel and Barany prepared N^α-Z-amino acids as 1,3-dioxolanes (one case as the dimethyl acetal) in four steps, loaded them onto BAL-PEG-PS by reductive amination, acylated the penultimate residue with HATU, and then ran routine Fmoc chemistry ^[2]. Guillaumie and colleagues worked out the Fmoc-compatible building block in detail. Reducing Fmoc-amino acid Weinreb amides with DIBAL-H at −78 °C worked but suffered partial loss of Fmoc; the cleaner sequence was isobutyl chloroformate mixed anhydride, NaBH₄ reduction to the amino alcohol, then Swern oxidation, giving Fmoc-Phe-H in 65% and Fmoc-Ala-H in 67%. The aldehyde was then converted to the dimethyl acetal with trimethyl orthoformate and catalytic TsOH in methanol. Removing Fmoc from the acetal proved awkward enough that an in-situ protocol — DBU in DMF, then acetic acid, then transfer to the aldehyde resin with NaBH₃CN — was preferred over isolating the amino acetal. Cleavage with TFA–H₂O (19:1) released the peptide and unmasked the acetal in the same step ^[3].

Two cautions come out of that work. Crude HPLC purities of the Fmoc-protected peptide aldehydes ranged from 50% to 84%, with the lowest values from dimethyl acetal precursors and the highest from 1,3-dioxolanes. And the cocktail matters: TFA–ethanedithiol (19:1) gave thioacetal adducts rather than the aldehyde ^[3]. A nucleophilic thiol or alcohol in the cocktail can compete for the carbonyl you have just liberated, but not always: in the same study the EDT-containing Reagent R cocktail released the aldehyde with no thioacetal formed. Konno's variant makes the point from the other side — direct 95% TFA treatment of their acetal-on-resin construct gave the peptide aldehyde in only 8% yield alongside 32% of an acetal carboxylic acid, which is why they instead converted the acetal to a thioacetal (EtSH, catalytic BF₃·Et₂O) and unmasked it with NBS in aqueous dichloromethane ^[1].

Oxazolidines from a resin-bound amino alcohol

An N,O-acetal formed between the aldehyde and a support-bound β-amino alcohol is an attractive mask because it can be broken under conditions much milder than a TFA cocktail. Ede and colleagues condensed Fmoc-Arg(Pbf)-H with resin-bound threonine to form the oxazolidine, elongated the chain, removed side-chain protection with TFA, and then released the aldehyde with 0.1% aqueous TFA in acetonitrile at 60 °C for 30 min. Across a 400-member library built from a 7 µmol loading of Fmoc-Arg(Pbf)-H, the average yield was 60%, and the mildness of the release step allowed the unpurified aldehydes to go straight into a protease assay ^[6].

The same oxazolidine chemistry serves for purification rather than anchoring. Sorg and co-workers oxidised peptide alcohols with a polymer-bound IBX derivative, then captured the aldehyde selectively on threonine-functionalised aminomethyl resin and released it afterwards, reporting excellent purity, satisfying yield, the optical integrity of the C-terminal residue conserved to a high degree, and compatibility with common side-chain protecting groups ^[7]. Read that stereochemical claim against the systematic study above ^[4]: "no epimerisation detected" generally means none detected by the analysis used, and a diastereomeric reference is what turns it into a measurement.

Weinreb amide reduction on the support

If the C-terminal residue is held as a Weinreb-type N-methoxyamide, hydride reduction stops at the aldehyde and cleaves it from the support in one operation. Gazal, Masterson and Barany built the resin-bound version: BAL-PEG-PS was treated with methoxylamine hydrochloride and NaBH₃CN to give a resin-bound methoxylamine, which was acylated with Fmoc-amino acids under PyBrOP activation and elongated normally. The same resin-bound Weinreb amide then diverged — TFA with scavengers gave the peptide hydroxamates in 68–83% crude yield and at least 85% initial purity, while LiAlH₄ in THF gave the C-terminal aldehydes in 16–53% crude yield and 30–40% initial purity ^[8]. That contrast, from one set of resins and sequences, is the clearest statement of what the aldehyde costs. Nor is the reduction perfectly chemoselective: under those LiAlH₄ conditions, model peptides carrying t-Bu-protected aspartate underwent partial side-chain reduction to homoserine-containing peptides ^[8].

Semicarbazone anchoring

A semicarbazone is a masked aldehyde and a linker at once. Siev and Semple anchored C-terminal aldehyde precursors to hydrazino-carbonyl-amino-methylated polystyrene through a semicarbazone linkage, elaborated the chain, and recovered the aldehydes by acidic hydrolysis in good overall yields and excellent purity ^[9]. The group being cleaved is the group being unmasked, so there is no separate deprotection to optimise.

Anchoring that generates the carbonyl at cleavage

Two strategies avoid carrying any carbonyl at all. The first is side-chain anchoring with a late on-resin oxidation. Page and colleagues attached Dde-tyrosinol through its phenol to a brominated Wang linker derivative, assembled the sequence by standard Fmoc chemistry with no evidence of esterification side reactions, and oxidised the resin-bound peptide alcohol with sulfur trioxide–pyridine; a short PEG spacer between linker and support dramatically improved the efficiency and reliability of that oxidation. Acidolysis then released the aldehyde ^[10]. This is ordinary side-chain anchoring, of the kind used to make C-terminal amides and hydrazides, with the difference that the anchored side chain is itself the masked carbonyl.

The second builds the carbonyl out of the linker itself. Melnyk and co-workers anchored a protected tartrate to PEGA or PEG-PS, ran Fmoc/tBu chemistry, deprotected both the peptide and the 1,2-diol on resin, and then treated the fully deprotected peptidyl-resin with periodate: the oxidation cleaved the product from the support and generated the C-terminal α-oxo aldehyde in the same step, with the spacing between carbonyl and peptide adjustable by linker design ^[11]. Note the analyte class — an α-oxo aldehyde (a glyoxylyl group) is not an α-amino aldehyde. It carries no stereocentre next to the carbonyl, so the epimerisation problem that dominates the routes above does not arise. Where the carbonyl is wanted for conjugation rather than active-site binding, that is a large simplification.

What is actually in the tube

The best-characterised case is leupeptin, an arginine aldehyde. By ¹H NMR in D₂O it exists as three interconverting covalent forms: the hydrate at 42%, a cyclic carbinolamine at 56% formed by addition of the guanidino ε-nitrogen to the aldehydic carbon, and the free aldehyde at just 2% ^[12]. The cyclic form is specific to the arginine side chain, but hydration is not: N-acetyl-L-phenylalaninal is likewise predominantly hydrated in aqueous solution ^[5], and in a 400-member library of C-terminal arginine aldehydes every mass spectrum showed an [M+H+H₂O]⁺ peak from partially hydrated material ^[6].

The chromatographic and mass-spectrometric consequences were dissected for the calpain inhibitor SJA6017. Its reverse-phase HPLC gives two peaks, each containing multiple species, and LC-MS/MS assigned four entities across them: the parent aldehyde, its methyl hemiacetal, the methyl enol ether, and the gem-diol. The hemiacetal forms when the amide-activated aldehyde reacts with methanol; that species dehydrates in a hot ion source to the enol ether, while both hemiacetal and gem-diol decompose back to the parent in the source. Hydration to the gem-diol happens during sample incubation and preparation, and hemiacetal and gem-diol are stable enough on the column to have distinct retention times ^[13].

This sets the limits of what your analysis licenses. An area-percent figure from a gradient RP-HPLC run on an aldehyde peptide is not a purity in the ordinary sense: two peaks may be the same compound in two hydration or solvation states, in a ratio that depends on the mobile phase and on how long the sample sat. Equally, a mass spectrum reporting M+18 is not evidence of an oxidised impurity, and one reporting only M+H is not evidence the material was dry — the adducts fall apart in the source. What the chromatogram cannot resolve for you is the epimer, a genuinely different compound that may co-elute. Diastereomeric purity has to be established against a synthesised reference of the opposite configuration ^[3] or by NMR on a model compound ^[4]; it does not fall out of the trace that shows the hydrate.

Handling and storage

Keep alcohols out of anything that touches the compound if you want one species. Methanol is a common substitute for DMSO in preparing stocks, and it is exactly what converts the aldehyde to a hemiacetal ^[13]; aqueous acetonitrile is the safer default. Purify immediately after unmasking: Konno's products epimerised at the α-carbon bearing the aldehyde and then decomposed, so quenching the oxidation quickly and going straight onto silica or RP-HPLC mattered ^[1]. Choose the cleavage cocktail with the carbonyl in mind, since a thiol scavenger can trap it as the thioacetal ^[3]. Where the route permits a mild release, crude material can go straight into an assay, avoiding a purification that would itself cost stereochemical purity ^{[4, 6]}. And store the masked precursor rather than the aldehyde whenever the schedule allows.

What the trouble buys

Chemoselective oxime and hydrazone conjugation

An aldehyde reacts with an aminooxy or hydrazide nucleophile at mildly acidic pH in the presence of every side chain a peptide has, which is what makes it a conjugation handle. The C-terminal α-oxo aldehydes made by the periodate route were shown to be useful partners in hydrazone, thiazolidine and oxime ligations ^[11]. Rates depend strongly on the carbonyl and on catalysis: with aniline catalysis and aromatic aldehydes, oxime and hydrazone ligations to 6-hydrazinopyridyl and aminooxyacetyl nucleophiles reach second-order rate constants of 10¹–10³ M⁻¹ s⁻¹ at ambient temperature and neutral pH, fast enough to label peptides and proteins at low micromolar concentration without a large excess of either partner ^[14]. Those numbers belong to aromatic aldehydes and should not be transferred to an aliphatic C-terminal α-amino aldehyde unchecked.

The linkage you form is reversible, and the choice between oxime and hydrazone is a choice about how reversible. Measured across pD 5.0–9.0, hydrolysis of both is acid-catalysed, and rate constants for oxime hydrolysis were nearly 10³-fold lower than those for simple hydrazones; a trialkylhydrazonium ion formed after condensation was more stable still ^[15]. If the conjugate has to survive a long incubation, that difference is the design decision. Oxime chemistry accepts a ketone as well as an aldehyde, and a C-terminal aldehyde is only one way to install either: N-terminal serine or threonine oxidation, transamination, engineered post-translational modification enzymes, unnatural amino acids and glycan modification all place an aldehyde or ketone site-specifically and feed the same aniline-catalysed chemistry ^[16]. Take the C-terminal route when the position, not just the presence, of the carbonyl is what you need.

Reversible covalent binding to proteases

The second use exploits the electrophilicity that makes synthesis hard. A peptide aldehyde matching a protease's substrate preference binds in the active site and the catalytic nucleophile adds to the carbonyl, giving a tetrahedral adduct that mimics the transition state but cannot proceed to acyl-enzyme. Cross-saturation NMR on N-acyl-L-phenylalaninals with α-chymotrypsin showed that although the compound is predominantly hydrated in solution, the aldehyde and not the hydrate is the effective inhibitor, and that the initial complex is in equilibrium with a hemiacetal formed between the aldehyde and the active-site serine ^[5]. Crystallography on the leupeptin–trypsin complex, solved at 1.7 and 1.8 Å in two crystal forms, put a length on the bond: the aldehyde carbonyl forms a covalent bond of 1.42 Å to the side-chain oxygen of Ser-195 ^[17].

The stereochemistry of that adduct is not settled between methods. In the crystal structures the hemiacetal oxygen points out of the oxyanion hole and hydrogen bonds to His-57 ^[17]. Solution and solid-state ¹³C NMR on leupeptin labelled at the argininal carbonyl instead resolves two interconverting diastereomers at the new tetrahedral centre — signals at δ 98.8 and δ 97.2 in an 84:16 ratio at pH 7, the minor one dominating at pH 3.0 — and assigns the major high-pH species as the S configuration, in which the hemiacetal oxygen sits in the oxyanion hole; its shift titrates with a pK_a of 4.69 ^[18]. If you are modelling an aldehyde adduct, the configuration at the new centre and its pH dependence are open questions, not fixed inputs.

The solution equilibrium also shows up in the kinetics, and it is easy to misread. Leupeptin binding to cathepsin B and papain shows a lag phase, which looks like slow binding. It is not: with only 2% of the material present as free aldehyde, and dissociation rate constants of 0.072 min⁻¹ for cathepsin B and 0.024 min⁻¹ for papain, the approach to steady state is slow even though association of the aldehyde form is fast, 1.2 × 10⁷ and 1.8 × 10⁷ M⁻¹ s⁻¹ respectively ^[12]. The practical reading is that an apparent K_i or association rate measured on total compound is wrong by whatever compound-specific fraction sits as hydrate or cyclic adduct at the pH of the assay. If you need a rate constant rather than a rank ordering, that distribution has to be measured, not assumed.