Peptide Synthesis

Why the order of operations is the whole difficulty

An amino acid has an amine at one end and a carboxylic acid at the other, and most have a reactive side chain as well. Put two in a flask with an activating agent and they will couple — in both directions, at both ends, and through whichever side chains happen to be nucleophilic. The product is a statistical mess.

Controlled synthesis therefore means masking everything that must not react, activating exactly one carboxyl, and unmasking one amine at a time. This is why a peptide chemist spends more effort on protecting groups than on bond formation. The governing requirement is orthogonality: the temporary group on the growing chain's N-terminus must come off under conditions that leave every side-chain group and the resin linkage untouched, cycle after cycle.

The second difficulty is separation. A failed coupling does not produce something obviously different — it produces the same peptide missing one residue. A 30-residue target and its 29-residue deletion sequence differ by a few percent in mass and often by almost nothing in retention time. This is why purification, not chemistry, is frequently the limiting step, and why yield and purity are different problems.

Solid-phase peptide synthesis: the cycle and its arithmetic

Merrifield's insight in 1963 was to anchor the growing chain to an insoluble support, so that purification between steps collapses into filtration and washing: reagents can be used in large excess and then simply rinsed away. The method earned the 1984 Nobel Prize in Chemistry and remains the default for peptides up to roughly 50 residues (Merrifield, Adv Enzymol 1969).

The chain is assembled from the C-terminus toward the N-terminus — the reverse of ribosomal synthesis. The C-terminal residue is attached to the resin, and each cycle then deprotects the N-terminus, couples the next protected amino acid, and washes. When the sequence is complete, a cleavage cocktail releases the peptide from the resin and removes the side-chain protecting groups, usually in a single step.

The arithmetic is unforgiving and worth internalising. Because every cycle multiplies, overall yield is the per-cycle efficiency raised to the number of couplings. At 99% per step, a 30-residue peptide returns about 74% of theory; at 98%, about 55%; at 95%, about 21%. A single percentage point of coupling efficiency is worth more on a long sequence than any downstream optimisation, which is why double couplings, excess reagent and elevated temperature are routine rather than exceptional.

Resin and linker

The linker decides the C-terminal chemistry before a single residue is added. Wang-type linkers release a C-terminal acid, Rink amide gives a C-terminal amide, and 2-chlorotrityl chloride resin can be cleaved under mild acid to release a fully side-chain-protected fragment — which is what makes it the resin of choice for fragment condensation.

Fmoc chemistry

The modern default. The N-terminal 9-fluorenylmethyloxycarbonyl group is removed with a base, typically piperidine, while side chains carry acid-labile protection. Base and acid are genuinely orthogonal, so no step threatens the others, and final cleavage uses TFA rather than hydrogen fluoride (Behrendt, J Pept Sci 2016).

Boc chemistry

The older strategy. The N-terminal group is acid-labile and side chains are benzyl-based, so deprotection and final cleavage differ only in acid strength — final release requires anhydrous HF. It survives because that same strong-acid regime suppresses aggregation, making it valuable for sequences Fmoc chemistry struggles with, and because it is the standard route to C-terminal thioesters for ligation.

Coupling reagents

The carboxyl must be activated to react. Carbodiimides such as DIC with an additive like Oxyma or HOBt are inexpensive and standard; uronium and phosphonium reagents such as HBTU, HATU and COMU are faster and more effective on hindered couplings. Additives are not optional garnish — they suppress the epimerisation that activation otherwise invites, which matters most at cysteine and histidine.

Where syntheses actually fail

Vendor literature describes SPPS as a repeating cycle. Practitioners know it as a series of specific, named failure modes, most of which are sequence-dependent and predictable in advance. Knowing them is most of what separates a route that works from one that does not (Coin, Nat Protoc 2007).

The common thread is that none of these produce an obviously broken product. They produce a closely related impurity that co-elutes with the target, which is exactly the impurity hardest to detect and hardest to remove.

On-resin aggregation

From roughly the sixth residue onward, chains can associate into β-sheets on the support. Reagents stop reaching the N-terminus, coupling and deprotection both slow, and yield collapses. Hydrophobic and alternating-hydrophobic sequences are the usual offenders. The classic fix is a pseudoproline dipeptide: a serine or threonine temporarily converted to a proline-like ring that kinks the backbone, breaks the sheet, and reverts to the native residue on cleavage — the technique that made amyloidogenic sequences such as human IAPP routinely accessible (Abedini, Org Lett 2005).

Aspartimide formation

The best-known sequence-specific side reaction in Fmoc chemistry. Aspartic acid, most severely when followed by glycine, cyclises onto the backbone under repeated base treatment. The ring then reopens to a mixture of the correct α-peptide, the β-isomer, and their D-epimers — several impurities of identical mass, so mass spectrometry alone will not reveal them. Mitigations include bulky ester protection, backbone protection, additives in the deprotection mixture, and shorter base exposure (Kong, ChemBioChem 2025).

Deletion and truncation sequences

An incomplete coupling leaves a free amine that couples in the next cycle instead, yielding a chain short by one residue. Capping with acetic anhydride converts those chains into truncated products that stop growing, which is preferable: a truncated peptide is easier to separate than a deletion sequence differing from the target by one residue in the middle.

Epimerisation

Activation makes the α-carbon acidic and invites loss of stereochemistry, with cysteine and histidine the most vulnerable residues. A D-epimer has the same formula and mass as the target and can co-elute. Additive choice, base choice and temperature are the controls.

Diketopiperazine formation

At the dipeptide stage the free N-terminus can attack the resin linkage and cyclise, cleaving the dipeptide from the support entirely. Proline and glycine at position two are the classic risk, and the standard avoidance is a bulkier linker or a different coupling order.

Ligation: how the length ceiling was broken

Stepwise SPPS degrades with length. Somewhere between 50 and 70 residues the yield arithmetic and the accumulating impurities make purification impractical. For decades that was the ceiling on chemical access to proteins.

Native chemical ligation removed it. Two unprotected synthetic fragments are joined in aqueous buffer: one carrying a C-terminal thioester, the other an N-terminal cysteine. The cysteine thiol attacks the thioester to form a transient thioester-linked intermediate, which then undergoes a spontaneous, irreversible S-to-N acyl shift to give a native amide bond at the junction. No protecting groups, no coupling reagents, and complete chemoselectivity even with other cysteines present (Dawson, Science 1994).

The method's obvious limitation — it needs a cysteine at every junction — was answered by desulfurisation, converting the junction cysteine to alanine afterwards, and by thiolated auxiliaries that mimic cysteine and are removed later. Together these put proteins of well over 100 residues within reach of total chemical synthesis, with full control over unnatural residues and site-specific modification (Kent, Chem Soc Rev 2009).

When biology is the better route

Chemistry scales badly with length and well with modification; biological expression is the reverse. Past roughly 50 residues, and for anything where every residue is one of the twenty canonical amino acids, recombinant production is usually cheaper and cleaner.

The target is normally expressed as part of a larger fusion protein — the fusion partner improves expression, keeps a toxic or protease-sensitive sequence out of trouble, and provides an affinity handle for capture. The peptide is then released enzymatically or chemically and purified. Insulin has been produced this way at industrial scale for decades.

The trade-off is what the cell can and cannot do. A ribosome cannot install a D-amino acid, an unnatural residue, a fatty-acid chain at a specific lysine, or a hydrocarbon staple. Most modern peptide drugs therefore use a hybrid route: recombinant expression for the backbone, chemistry for the modifications that give the molecule its properties.

E. coli

Fast, inexpensive, very high yields. No native machinery for eukaryotic post-translational modification, and expressed product often needs refolding from inclusion bodies.

Yeast (Pichia, Saccharomyces)

Secretes product into the medium, simplifying capture, and performs disulfide formation and glycosylation — though with a glycan pattern that differs from the human one.

Mammalian cell lines

The most faithful post-translational modification and the most expensive. Reserved for targets where authentic modification is part of the activity.

Cell-free systems

Transcription and translation in a lysate rather than a living cell. Slower to scale, but tolerates toxic products and allows unnatural amino acids through engineered translation components.

Purification, and what a purity figure does not tell you

Crude peptide is a mixture of the target and its close relatives. The standard separation is preparative reversed-phase HPLC on a C18 stationary phase with a water–acetonitrile gradient and an ion-pairing acid, a method that has been the backbone of the field for forty years (Rivier, J Chromatogr 1984). Purification is routinely the most expensive step in the whole process, and on a difficult sequence it can discard most of the material.

Identity and purity are then established separately, and the distinction matters. Analytical HPLC integrates peak area to give a purity percentage — how much of what was injected came off as one peak. Mass spectrometry establishes that the peak is the right molecule. Neither answers on its own: a purity figure without a mass confirms a clean something, and a mass without purity confirms the right molecule is present somewhere in the mixture.

A third number is the one most often missing. Net peptide content is the fraction of the powder's mass that is peptide at all, the remainder being counter-ion — usually acetate or TFA salt — and bound water. A preparation can honestly be 99% pure and 80% peptide content simultaneously, which means a vial labelled 10 mg holds about 8 mg of peptide and every concentration calculated from the label is high by a fifth.

Scale, cost and the solvent problem

SPPS is chemically elegant and materially wasteful. Reagents are used in large excess by design, and each cycle is followed by several resin washes, so solvent dominates the mass balance. Process mass intensity — kilograms of input per kilogram of product — runs far higher for peptides than for small-molecule drugs, and the great majority of it is solvent rather than reagent (Kekessie, J Org Chem 2024).

This became a commercial problem rather than an aesthetic one for two reasons. Demand for peptide APIs rose sharply with the incretin drugs, putting manufacturing capacity and cost under real pressure. And the field's default solvents, DMF and NMP, are reprotoxic and under tightening regulatory restriction, which turns solvent choice into a supply question rather than a preference (Isidro-Llobet, J Org Chem 2019).

Replacement work is therefore active and practical. N-butylpyrrolidinone has been reported as both greener and synthetically better than DMF in SPPS (Kumar, ChemSusChem 2020), alongside other candidate solvents and a broader industrial effort to cut waste through reagent recovery, higher loading and shorter washes (Martin, RSC Adv 2020). For a buyer, the relevant consequence is that solvent strategy now shows up in what a contract manufacturer can quote and deliver.

Automation, microwaves and flow

Automated synthesisers have handled the repetitive cycle for decades. Two developments have changed what is achievable rather than merely convenient.

Microwave-assisted SPPS applies controlled heating to coupling and deprotection, cutting step times and improving conversion on sequences where aggregation slows diffusion — with the caveat that heat also accelerates aspartimide formation and epimerisation, so it is a tool to be aimed rather than applied uniformly (Singh, Methods Mol Biol 2020).

Automated flow chemistry goes further: reagents are pumped continuously through a heated packed bed rather than batched in a vessel, collapsing each cycle from tens of minutes to a small number of minutes and making the chemical synthesis of protein-length chains a practical exercise rather than a heroic one (Hartrampf, Science 2020).

Choosing a route

In practice the decision turns on three questions: how long is it, what is attached to it, and how much is needed. The table below is a first approximation rather than a rule — a difficult 25-mer can be harder than a routine 45-mer.