A pseudoproline is a ring you build across a serine, threonine or cysteine residue and then take apart again at cleavage. Condensing the side-chain hydroxyl or thiol onto a carbonyl carbon closes a five-membered oxazolidine or thiazolidine that ties the side chain back to the backbone nitrogen. That nitrogen now sits inside a ring and carries no hydrogen, so for the duration of the assembly the residue behaves like a proline: it cannot donate a backbone hydrogen bond, and the amide immediately N-terminal to it is biased into a conformation a beta-sheet cannot accommodate [2, 3]. At the acid cleavage step the ring opens and the native residue is regenerated, leaving nothing to remove in a separate operation. The building blocks were introduced for standard Fmoc/tBu synthesis and incorporated by coupling preformed, suitably protected dipeptides; the original report demonstrated their potential to solubilize otherwise sparingly or completely insoluble model peptides and described them as having an intrinsic propensity for preventing peptide aggregation and beta-sheet formation [1].
This guide is about the chemistry of that ring and of the benzyl-type alternatives that do the same job on a nitrogen with no side chain to work with. It assumes you already know why a chain aggregates on resin and what a deletion sequence looks like in the crude.
The ring, and why the gem-dimethyl group is load-bearing
Commercially available Fmoc pseudoproline dipeptides consist of five-membered oxazolidine or thiazolidine rings synthesized through acid-catalyzed reaction of serine, threonine or cysteine with 2,2-dimethoxypropane. The resulting acetonide acts as an acid-labile protecting group for both the beta-hydroxyl or thiol group and the Nalpha-amide at the same time — one operation, two protections. The two methyl substituents at the ring 2-position are essential for that acid lability: unsubstituted analogues are TFA-stable [2]. The psi(Me,Me)pro on the label is not just nomenclature; without those two methyls the ring would survive the cleavage instead of opening. A cysteine derivative in common use is built differently — Cys(psiDmp,Hpro), carrying a 2,4-dimethoxyphenyl substituent and a hydrogen at the ring 2-position rather than two methyls [9].
What the kink actually is
Two distinct things happen when the ring closes, and they are worth separating because only one of them is a measured quantity.
The first is the loss of a donor. A tertiary amide has no N-H, so one node is deleted from the hydrogen-bond network that a beta-sheet is built from. The second is conformational. Disruption of secondary structure by these building blocks is attributed to the cisoid amide conformation of the oxazolidine- or thiazolidine-based ring, which introduces a kink in the backbone and disrupts backbone hydrogen bonding analogously to proline [3]. Pseudoprolines in peptides favour a cis-amide conformation, with a cis/trans ratio of approximately 95:5, attributed to a steric clash between the substituent or substituents at the ring C2 position and the side chain of the N-terminal vicinal residue when the amide is trans [2].
Note what that mechanism implies. The amide whose geometry is biased is the one preceding the pseudoproline, and the residue on the other side of that amide supplies half of the steric clash. The identity of the neighbour is part of the mechanism, not an incidental detail of how the building block is sold.
The same turn-inducing effect is exploited deliberately outside difficult-sequence work. Incorporating a pseudoproline into a linear precursor generally results in a cyclization that proceeds more quickly and with higher yield than the analogous sequence without it, and installation of a pseudoproline at the C-terminal position of a linear sequence has been shown to eliminate epimerization of that residue during the reaction [4]. Both observations are reported for cyclization chemistry on peptides carrying serine-, threonine- or cysteine-derived rings; they are not a general statement that any pseudoproline improves any coupling.
Placement: not at the residue that failed
The instinct is to put the ring where the coupling broke down. That is usually the wrong place, because the structure responsible was built from residues already installed. Backbone protection is described as particularly effective when introduced at approximately every six residues [2] — a spacing rule, not a targeting rule.
The reach of one aggregating residue is longer than it looks. Writing about difficult peptides in general, the authors of an automated study of a 38-residue sequence with a difficult pattern report that hydroxyl amino acids such as serine can cause severe breakdowns in coupling yields even several amino acids after the insertion of the critical amino acid; that study itself compared resins, activators and a structure-breaking pseudoproline dipeptide alone and in combination [5]. One correctly placed ring therefore covers a window rather than a position.
How narrow that window is has been documented directly. For a 54-residue caveolin-1 fragment encompassing the intramembrane domain, a slight modification in the positioning choice conditioned whether the synthesis was achieved at all; the same work reports a side reaction correlated with the coupling steps that generated truncated fragments with a mass deviation of +42 Da [6]. Count is also a variable. For a chaperonin 60.1 sequence described in the literature as inaccessible, two syntheses using one and two pseudoproline units respectively were compared with an ordinary stepwise double-coupling procedure, and only the synthesis combining two pseudoproline building blocks succeeded [7]. At the far end of the scale, judicious use of dimethyloxazolidine dipeptides of serine and threonine allowed a 95-residue fragment related to the FAS death domain to be prepared with single one-hour couplings, after attempts with conventional building blocks had failed [8].
Those are four sequences, not an algorithm. What they license is a working procedure: identify the serine, threonine and cysteine positions available in the difficult stretch, place rings at roughly six-residue intervals across it, and expect to need more than one. What they do not license is predicting from sequence alone which single position will rescue a given synthesis.
Placement matters for a second reason that has nothing to do with aggregation. Aspartimide formation at Asp-Ser, Asp-Thr and Asp-Cys junctions can be avoided by using pseudoproline dipeptides, and where a pseudoproline cannot be placed vicinal to the aspartyl residue on its C-terminal side, one introduced a residue earlier has been used to induce a backbone conformation unfavourable for aspartimide formation [2].
Why they are sold as dipeptides
The ring nitrogen you have just created is a hindered secondary amine, and the next residue in the assembly has to acylate it on resin. Because of the difficulty of acylating that sterically hindered secondary amine, the derivatives are often used in the form of Yyy-Xxx(psiPro) dipeptides [9] — the hard bond is formed in solution by the supplier, and what you couple on resin is an ordinary activated carboxyl.
Do not read that as a prohibition. Fmoc-protected pseudoprolines have been synthesized and applied in peptide synthesis not as dipeptides but as individual amino acids, with different acylation protocols and amino acids tested and several difficult peptides made to confirm that the constructions work in automated or manual synthesis [10]. In a 16-residue peptide made up of the C-terminal sequence of human growth hormone plus an extra N-terminal tyrosine, two serines were introduced as the single amino acid Fmoc-Ser(psiMe,Mepro)-OH, which the authors present as demonstrating that acylation of these hindered moieties is possible and as a way to avoid the corresponding dipeptides, which are very often not commercially available or very costly [11]. The review position is consistent: monomers can be introduced, with double coupling or flow chemistry recommended for the subsequent acylation [2].
So the dipeptide is a convenience you pay for twice: in price, and in losing a free choice of the neighbouring residue. The monomer is cheaper and more flexible and moves the hindered coupling back onto your resin, where it becomes your problem to drive.
Regeneration at cleavage, and the failure mode to look for
Oxazolidine rings derived from serine and threonine are typically deprotected to the parent amino acid within several hours in TFA-based cleavage cocktails [3]. Thiazolidines have a more complicated record, and it is a good example of a limitation being narrower than the literature first made it look. Deprotection times of 32 hours were reported in the seminal work for removal of cysteine pseudoprolines in linear peptides, and a later publication required 36 hours in TFA/TIS/H2O (95:2.5:2.5) for a linear peptide and reported that head-to-tail cyclic peptides took much longer and needed harsh acids. Against that background, Postma and Albericio report a series of peptides in which complete removal took two hours or less at room temperature, with two peptides requiring 45 degrees Celsius and two to three hours, and conclude that cysteine pseudoproline removal is sequence-dependent and that the governing factors are not well understood [3]. Treat the deprotection time as something to verify on your own sequence, not as a property of the building block.
The specific artefact to look for in the crude is a peptide carrying the ring still closed. In continuous-flow synthesis of a model decapeptide, a truncated pentapeptide was recovered whose measured mass was approximately 40 g/mol above the expected value; the intact pseudoproline oxazolidine ring was identified as responsible, and it survived 95% TFA. Batch to batch, that species reached 2 to 7% of the crude peptide. The same study also found aspartimide formation to be catalyzed rather than hindered by the pseudoproline moiety, and attributes both side reactions to elevated temperature and pressure, advising caution when using these derivatives under such harsh conditions [9]. That is one model sequence under flow conditions, and the authors frame it as a warning about the conditions rather than a re-evaluation of the building block; but it means a mass search of your crude should include the closed-ring species, and it means the routine claim that pseudoprolines suppress aspartimide has at least one documented counter-case.
When the sequence offers no hydroxyl or thiol
Serine, threonine and cysteine together represent under 14% of amino acid natural abundance, which is the structural ceiling on the whole approach [2]. The stretches most likely to aggregate — aliphatic-rich segments — are precisely the ones least likely to supply a usable position. The alternative is to put a removable substituent directly on a backbone nitrogen.
The 2-hydroxy-4-methoxybenzyl (Hmb) group is the reference case. Introducing it installs a reversible protecting group on the backbone amide bonds so that aggregation due to hydrogen-bond interchain association is inhibited [12]. It is efficiently introduced into most Nalpha-amides, although incorporation at junctions with flanking beta-branched side chains is challenging, and it can be installed in situ by reductive amination of the corresponding benzaldehyde onto the N-terminal amino group [2]. The reason it works at all on a hindered secondary amine is the 2-hydroxyl: the incoming Fmoc-amino acid is captured as an ester on the phenol and then delivered by an O-to-N acyl shift to give the tertiary amide [2]. For the acyl carrier protein decapeptide, the final valine addition is typically 10 to 15% incomplete under conventional Fmoc conditions, and with Hmb protection at Ala68 that coupling proceeds to completion [2].
The costs are specific. In a direct comparison of Hmb-protected amino acids against pseudoproline building blocks in the solid-phase synthesis of difficult peptides, both gave substantial improvements in crude purity, but pseudoproline incorporation was found to be superior to Hmb backbone protection, and the authors attribute that to slow and incomplete coupling of the amino acid immediately following the Hmb amino acid [13]. Position sensitivity is documented too: in analogues of a beta-amyloid C-terminal fragment monitored by on-line Fmoc deprotection profiles, hindered release was still seen in the (Hmb)Gly38 analogue as in the unmodified peptide, no such hindrance was observed for (Hmb)Gly37 or (Hmb)Val36, and the (Hmb)Val36 synthesis nonetheless contained an exceptionally slow coupling reaction [14]. One position removed the structural problem; a neighbouring position did not; and a third removed it while creating a kinetic one.
Scale changes the arithmetic again. Hmb is described as poorly suited to large-scale synthesis, where only 1.5 to 2.0 equivalents of amino acid are typically used, because the aryl 2-hydroxyl competes with the N-terminus during coupling and depletes the amino acid reactant; at the 5 to 10 equivalents typical of small-scale work this does not bite [2]. Practical detail on the building blocks themselves: monoFmoc Hmb amino acid derivatives were obtained in higher yield than the bisFmoc derivatives, and coupling yields to the amino peptide resin were the same for both under TBTU/HOBt/DIEA conditions [12].
The 2,4-dimethoxybenzyl (Dmb) group trades the same feature in the opposite direction. It has no reactive phenolic group to compete with the N-terminal amine during acylation, which is an advantage over Hmb, but for the same reason it cannot perform O-acyl capture, and it is therefore largely limited to N-protection of glycine; acylation of a Dmb-protected N-terminus can be improved by microwave heating or bypassed by using Dmb-containing dipeptides with the tertiary amide already formed [2]. The scale on which that works is substantial: the 61-residue glycine- and arginine-rich C-terminal domain of human nucleolin was assembled using the backbone-protected dipeptide Fmoc-Gly-(Dmob)Gly-OH, giving both the unmodified and the dimethylated peptides in approximately 26% yield and high purity [15], a synthesis the recent review describes as using 14 such building blocks and as improving HPLC yield from 5% to 26% [2].
Benzyl-type protection carries two liabilities of its own. Acid lability depends on the flanking side chains and is sequence-dependent, so incomplete deprotection is possible; and the benzylic cations liberated during cleavage can alkylate cysteine, methionine and tryptophan, which is why using numerous benzyl groups in a single synthesis is not recommended [2]. That is the argument for the dicyclopropylmethyl (Dcpm) alternative, introduced onto amino acids by reaction of dicyclopropylmethanimine hydrochloride with an amino acid ester followed by cyanoborohydride or triacetoxyborohydride reduction, and demonstrated on the alanine decapeptide and on the prion peptide 106-126, with Fmoc-(Dcpm)Gly-OH substituting for Fmoc-Gly-OH to avoid cyclization to the aminosuccinimide in Asp-Gly sequences [16]. On that prion fragment the review reports 41% crude purity with Dcpm against 7% for the corresponding Hmb synthesis [2] — a single-sequence comparison, and one that says nothing about how either group would perform elsewhere.
What is unresolved
Three things should stay open in your planning. The six-residue spacing figure is a review's summary of the primary literature, and the demonstrations behind it are individual sequences optimized one at a time; the caveolin and chaperonin cases show that shifting one position or adding one unit can decide the outcome [6, 7]. Cysteine thiazolidine deprotection ranges over more than an order of magnitude in reported time and the controlling factors are stated by the authors who narrowed that range to be not well understood [3]. And the interaction between pseudoprolines and aspartimide runs in both directions in the current record — suppression at Asp-Ser, Asp-Thr and Asp-Cys junctions [2], catalysis in one flow synthesis at elevated temperature and pressure [9]. None of these is a reason to avoid the chemistry; all of them are reasons to build the mass-spectrometric check for the closed-ring and aspartimide species into the analysis rather than assuming clean regeneration.