Aspartimide Side Reactions in Fmoc Peptide Synthesis

Every Fmoc removal cycle exposes the growing chain to a secondary amine in DMF for several minutes, and each exposure is an opportunity for an aspartyl residue to close on its own backbone. The nucleophile is the amide nitrogen immediately C-terminal to the Asp; the electrophile is the carbonyl of the protected β-carboxyl on the Asp side chain. Deprotonation of that backbone amide is followed by nucleophilic attack at the β-carbonyl, expelling the side-chain ester and leaving a five-membered cyclic imide fused into the backbone ^[1]. The same cyclization is described for asparagine as well as aspartate side chains ^[2].

Two features separate this from the epimerization that happens at the coupling step. It is driven by the deprotection base rather than by carboxyl activation, so none of the standard anti-racemization measures — omitting preactivation, switching to a hindered base, cooling the coupling — touch it. And it is cumulative in a way a coupling side reaction is not: an Asp installed early in a 40-mer is carried through every subsequent piperidine treatment, and each one is another chance to cyclize.

What one aspartimide becomes

The imide is a branch point, not an end product. It is electrophilic at both ring carbonyls, and whichever nucleophile reaches it first decides what you isolate.

Water opens it to a mixture. Cisse and Aucagne describe aspartimide as susceptible to nucleophilic attack by water leading to the formation of Asp and isoAsp, and note that the imide can also epimerize because of the increased acidity of the Cα hydrogen, which gives D-aspartimide, which in turn generates D-Asp and D-isoAsp ^[2]. A single cyclization event therefore feeds four ring-opened residues: the correct L-α-aspartyl linkage, the L-β-linked isoaspartyl chain, and the D forms of both. Kong and colleagues report that the D/L-β-aspartyl peptides are typically formed in a 3:1 ratio compared with the α-aspartyl peptide ^[1]. That ratio is the number worth remembering: the β-linked material is usually the majority product of ring opening, not a trace.

The deprotection base is itself a nucleophile. Piperidine, or any other nucleophilic base, reacts with the aspartimide to form α- and β-base adducts ^[1]. These piperidides are covalent adducts that persist to the finished crude: Lauer, Fields and Fields characterized and quantitated the α- and β-piperidides in cleaved peptide by RP-HPLC and mass spectrometry ^[5].

Any nucleophile in contact with a surviving imide will do the same job. Stathopoulos and colleagues opened aspartimide-containing peptides deliberately, testing primary and secondary alcohols against a panel of bases, and obtained α- and β-aspartyl peptide esters; their best results came from a primary alcohol with diisopropylethylamine, while ring opening by secondary alcohols occurred only at high temperatures ^[3]. Read backwards, that is a warning about incidental nucleophiles meeting material that reached the analysis with the ring intact: the same authors note that hydrolysis of aspartimide-containing peptides, especially during HPLC purification, yields an α/β mixture that cannot easily be purified, and their alcohol results show what a basic methanolic workup adds on top of that ^[3].

There is also a chain-termination route. Samson and colleagues observed significant amounts of chain termination in Fmoc-SPPS while synthesizing the N-terminal Xaa-Asp-Yaa motif, caused by the formation of piperazine-2,5-diones; NMR of the side product proved that exclusively the six-membered ring is formed and not the conceivable seven-membered 1,4-diazepine-2,5-dione, and they proposed a mechanism in which the N-terminal amino function attacks the α-carbonyl of the aspartimide intermediate ^[4]. That failure mode needs the Asp at the second position from the free N-terminus, but where it applies it converts a purification problem into a yield problem.

Sequence dependence

Most of the published ranking comes from one scaffold, H-Val-Lys-Asp-Xaa-Tyr-Ile-OH, with Xaa varied. Lauer, Fields and Fields reported that aspartimide formation occurred for X = Arg(Pmc), Asn(Trt), Asp(OtBu), Cys(Acm), Gly, Ser, Thr and Thr(tBu) ^[5]. Mergler and colleagues, working the same model systematically, found considerable amounts of by-products for Xaa = Asp(OtBu), Arg(Pbf), Asn(Mtt), Cys(Acm) and unprotected Thr ^[6].

Asp-Gly is the canonical worst case. Kong and colleagues state that the formation is especially prevalent in Asp-Gly motifs, likely because Gly is the least sterically hindered residue, and name Ala, Ser, Thr, Cys, Arg, Asp and Asn as the other C-terminal partners that make especially aspartimide-prone motifs ^[1]. The steric account is offered as a likely explanation rather than a measured barrier, and it does not cover everything, which is why the lists above carry protecting groups: the substituents on the following residue are part of the result.

Two cautions on reading those tables. They come from a resin-bound hexapeptide under deliberately extended base exposure, so the absolute percentages do not transfer to your sequence; what transfers is the ordering and the identity of the hazardous motifs. And the Asp side-chain protecting group is part of the system being ranked, not a constant. Ruczyński and colleagues protected Asp with the Dmab group across an Asp-Ala motif in a 15-residue galanin fragment analogue and reported a strong tendency of the Dmab-protected peptide to form aspartimide with unusual high efficiency ^[7]. Asp-Ala sits well down the motif list; the protecting group moved it to the top.

What RP-HPLC and MS each license you to conclude

Start with the mass arithmetic, because it decides the analytical strategy. Cyclization closes a ring with loss of water relative to the free acid, so the intact aspartimide is the M−18 species, and Cisse and Aucagne track it by exactly that label ^[2]. It is the one member of the family that a mass spectrum finds without effort.

The ring-opened products are where mass stops helping. Cisse and Aucagne state that three of the byproducts are isobaric to the target, that is, of identical molecular mass: isoAsp, D-Asp and D-isoAsp ^[2]. Isomerization moves a bond and epimerization moves a proton; neither changes the molecular formula. A correct intact mass on a peptide containing an Asp-Gly motif therefore excludes the imide and the base adducts and says nothing at all about the α/β ratio or the D content.

Base adducts run the other way on the mass axis. A piperidide is the imide plus one molecule of the amine, so it sits at M−18 plus the 85 Da of piperidine, nominally 67 Da above the target. Because the α and β adducts are regioisomers ^[1], one peak at that mass can be two compounds, and a shoulder on it is not necessarily a second impurity class.

Chromatography carries what mass cannot, and it degrades with chain length. The model-peptide studies determine the extent of aspartimide and related by-product formation by RP-HPLC ^{[6, 8]}, so on a hexapeptide the by-products are separated well enough to quantify. Cisse and Aucagne note that separation by standard HPLC of the target from aspartimide and from all the aspartimide-derived byproducts is not expected in the case of a long polypeptide chain, and argue that the prevalence of the side reaction may have been overlooked because of that analytical difficulty combined with in situ conversion of aspartimide into other byproducts having the same molecular mass as the parent aspartate residue ^[2].

If the question is where an isoAsp sits rather than whether one exists, conformation-selective methods are one route. Gibson and colleagues ran a cyclic ion mobility mass spectrometry system on a peptide and three synthetic isomeric forms, each with a single Asp residue isomerized to isoAsp, and correctly located which Asp residue in each peptide had isomerized ^[9]. Read the scope carefully: one peptide, with authentic synthetic isomers already in hand for comparison, and the authors present the technique as a potential screening tool rather than a validated general assay. Without those standards the fallback is co-injection against a deliberately synthesized isoAsp analogue.

Suppression 1: make the β-ester harder to reach

The electrophile is the side-chain ester carbonyl, so steric bulk around it slows the cyclization. Mergler and colleagues screened a set of β-carboxy protecting groups on the hexapeptide model and found that 3-methylpent-3-yl (Mpe) protection of the Asp side chain resulted in significant improvements with respect to aspartimide formation ^[8]. A later derivative, the β-(2,3,4-trimethyl-pent-3-yl) ester, reduced by-product formation more efficiently than the less bulky Asp(OMpe); both were superior to the standard Asp(OtBu), and the additional margin showed up specifically in cases of prolonged contact with piperidine or DBU, which the authors read as valuable especially in syntheses of long peptides or difficult sequences ^[10].

Behrendt and colleagues extended the idea to trialkylcarbinol esters — OEpe (3-ethyl-3-pentyl), OPhp (4-n-propyl-4-heptyl) and OBno (5-n-butyl-5-nonyl) — and compared them with OtBu and OMpe in scorpion toxin II model variants under piperidine and DBU at both room and elevated temperatures, reporting the new derivatives as extremely effective in minimising aspartimide by-products ^[11]. Applied to Asp-Gly-containing peptides and to a GLP-2 analogue whose synthesis is made challenging by extensive aspartimide formation, the same approach gave almost complete reduction of aspartimide formation with accompanied suppression of aspartic acid epimerisation ^[12]. That second clause matters for anyone reading a chiral assay: at an Asp residue, D-content and aspartimide content are not independent readouts.

Bulk has a ceiling. Kong and colleagues put the OBno result on that same GLP-2 analogue at a 25% reduction in aspartimide relative to Asp(OtBu), and note that limitations are reached when attempting to use even bulkier esters ^[1]. The gap between that figure and the almost-complete reduction claimed in the originating report is a reminder that suppression percentages are properties of the exact comparison being run, not specifications for a building block. Steric protection slows the cyclization; it leaves the electrophile in place.

Suppression 2: remove the nucleophile

Blocking the backbone amide nitrogen removes the nucleophile, which is the only measure in this list that is mechanistically complete rather than merely slower. The nitrogen to mask is the one belonging to the residue C-terminal to the Asp ^[2]. Mergler and colleagues achieved complete suppression using the combination of OtBu side-chain protection and Hmb backbone protection on the Gly of H-Val-Lys-Asp-Gly-Tyr-Ile-OH ^[8]. Kong and colleagues put numbers on the same comparison: with Hmb the toxin II fragment was obtained in 90% yield, whereas without it the fragment could only be isolated in low yield, and just 30% of that material was unmodified peptide, the remainder being aspartimide plus α- and β-piperidides ^[1].

The cost is acylating a hindered secondary amine. Abdel-Aal and colleagues addressed that with 2-hydroxy-4-methoxy-5-nitrobenzyl (Hmnb), introduced during assembly by adding 4-methoxy-5-nitrosalicylaldehyde to the peptidyl-resin followed by on-resin reduction to the secondary amine; acylation is aided by formation of an internal nitrophenol ester that undergoes a favourable O,N intramolecular acyl transfer, and they report that this activated ester generally gives complete reaction with standard coupling conditions ^[13]. They demonstrated it both on an aspartimide-prone test sequence and on classic aggregation-prone sequences, which is the second dividend of backbone protection: the same masked amide that cannot cyclize also cannot hydrogen-bond into a sheet.

Backbone protection is not confined to Fmoc assembly. Cisse and Aucagne applied a 2-(4-aminobutanoyloxy)-4-methoxybenzyl (GABA-Hmb) group to suppress aspartimide during native chemical ligation, where the imide forms in aqueous buffer rather than in piperidine, and detected no trace of the M−18 byproduct with the Hmb-based strategy against approximately 15% without it ^[2]. Those are ligation conditions and the numbers do not transfer to a deprotection cycle, but the mechanism and the countermeasure do.

Suppression 3: get rid of the ester entirely

Neumann and colleagues replaced the β-ester with a cyanosulfurylide, masking the carboxylic acid through a stable C–C bond instead of a C–O bond, so there is no ester carbonyl for the backbone amide to attack. They describe these formally zwitterionic groups as exceptionally stable to all common manipulations and as imparting improved solubility during synthesis, with deprotection achieved readily and rapidly under aqueous conditions using electrophilic halogenating agents via a highly selective C–C bond cleavage reaction ^[14]. They applied it to a range of peptides and proteins and state that the strategy has the potential to overcome one of the most difficult aspects of modern peptide chemistry — a stated potential, not a demonstrated universal result.

Suppression 4: change the deprotection mixture

The cheapest interventions act on the base cocktail itself. Michels and colleagues demonstrated that addition of small amounts of organic acids to the standard Fmoc cleavage agent piperidine efficiently prevents formation of aspartimide side products, and that this effect is virtually independent of the acid strength ^[15]. The independence from acid strength argues against any mechanism specific to one additive and in favour of simply moderating the medium.

Kong and colleagues collect the working numbers: 0.1 M hydroxybenzotriazole (HOBt) or 2,4-dinitrophenol reduced aspartimide formation significantly; ethyl 2-cyano-2-(hydroxyimino)acetate (Oxyma Pure) has been employed and has proven more effective than HOBt; and 5% formic acid was sufficient to reduce aspartimide formation by 90% in one reported synthesis ^[1].

Two limits on this class of fix. Additives moderate the deprotection medium, they do not remove the amine, so suppression is partial by construction — and a side reaction knocked down by 90% but given thirty opportunities still leaves a measurable population. Mergler and colleagues arrived at the same practical position from the other direction, reporting that aspartimide formation could be diminished by employing milder methods for Fmoc cleavage, such as hexamethyleneimine/N-methylpyrrolidine/HOBt/NMP/DMSO 4:50:4:71:71 (v/v/w/v/v) ^[6]. Diminished, not eliminated.

Exposure time and temperature are variables you set

The cyclization accumulates with base exposure rather than completing inside a single short deprotection, which is why the studies that stress-test protecting groups do it with prolonged or repeated base treatments, and why total base contact time is a lever. The advantage of the more hindered ester appeared in Mergler's data specifically under prolonged contact with piperidine or DBU ^[10], and Behrendt and colleagues evaluated their building blocks by treating peptidyl resins with piperidine- and DBU-containing reagents at both room and elevated temperatures ^[11]. Cisse and Aucagne measured the temperature dependence directly, though in aqueous ligation buffer rather than in DMF: dropping from 50 °C to 37 °C reduced the total amount of byproducts by nearly threefold, and by more than twentyfold at 20 °C, and phosphate buffer significantly increased both epimerization and aspartimide formation compared with HEPES ^[2]. Those are solution-phase ligation figures and should not be quoted as SPPS numbers; what they establish is how steep the temperature dependence can be, which is the reason elevated-temperature protocols deserve an explicit check at every Asp.

Why this is not the racemization problem

D-Asp appears in both discussions and the overlap misleads. Epimerization at the coupling step proceeds through an activated carboxyl and responds to activation chemistry, base identity and preactivation time. The D-Asp described here is generated afterwards, from a neutral cyclic imide whose Cα proton is acidified by the two flanking ring carbonyls, and in Fmoc chemistry that imide is formed during Fmoc removal ^{[1, 2]}. A protocol tuned to suppress coupling-step epimerization at Cys or His does nothing about it, because the offending step is not the coupling. The diagnostic is whether D-content tracks aspartimide content: Behrendt and colleagues found that suppressing aspartimide brought accompanied suppression of aspartic acid epimerisation ^[12], which is what you expect when the imide is the dominant route to inversion at that residue.

What to check

If the sequence contains Asp-Gly, Asp-Asn, Asp-Asp, Asp-Arg, Asp-Ser, Asp-Thr, Asp-Cys or Asp-Ala, treat aspartimide as the default hypothesis for any unexplained shoulder, and treat a correct intact mass as evidence about the imide and the adducts only. Three questions have to be separated. Is there an M−18 species — mass will find it. Is there a species around 67 Da above the target — mass will find that too, though it cannot tell the α adduct from the β. And what fraction of the correct-mass material is β-linked or D-configured — mass will not find that at all, and a single symmetric RP-HPLC peak does not exclude it, particularly on a long chain ^[2]. Answering the third needs chromatography developed against authentic isoAsp and D-Asp standards, or a conformation-selective method run alongside them ^[9]. The order of operations that costs least is to design the imide out during assembly rather than to chase its products afterwards.