Stepwise solid-phase assembly extends a chain one residue at a time on a support whose local environment gets worse with every addition. Incomplete couplings accumulate as deletion sequences that differ from the target by a single residue and therefore chromatograph almost exactly like it. Underneath that, the chain begins to fold against itself: aggregation of the growing peptide chain is the main cause of failure in peptide chemosynthesis [2]. The result is a working length window rather than a hard wall. Solid-phase peptide synthesis made peptides below roughly 50 residues broadly accessible, and it was the limitations of that technique for producing large homogeneous peptides that stimulated the development of solid-phase capture purification methods and other routes to large polypeptides [1].
Inside the window the method is fast — the 41-residue peptide CRF was obtained in about 80 working hours under one laboratory's standard Fmoc/tBu protocol on PEG-modified polystyrene [2]. Outside it, more effort at the same route does not always help. For one extremely difficult sequence, an Asn15 analogue of the WW domain FBP28, the standard protocol does not give the peptide at all, and depsipeptide and pseudoproline units had to be built into the chain to obtain it [2].
What a convergent route actually changes
Convergent peptide synthesis assembles the target from solid-phase-synthesised, suitably protected fragments covering the whole sequence, which are then condensed either on a solid support or in solution [3]. The argument for it is not primarily yield. It is that the failure product changes identity. In a stepwise synthesis the dominant impurity is short by one residue; in a fragment condensation it is short by an entire fragment, and target and side-product then differ enough physicochemically to be separated [5]. The stated cost of the approach, in the same source, is the unpredictable solubility and reactivity of protected fragments [5].
Convergence is also divisible. A single fragment condensation can be inserted into an otherwise conventional stepwise synthesis purely to get across a difficult region [5]. Where the whole sequence resists, the fragmentation itself becomes the design variable: the peptide TD2.2 was obtained by sequential condensation of four protected fragments after several unsuccessful attempts by the stepwise approach, with the fragments chosen to minimise the number of demanding amino acids in each, so that hydrophobic residues and Arg(Pbf) were spread over multiple fragments rather than congested in one [12].
Preparing the protected fragments
A protected fragment has to come off its resin with every side-chain and Nα protecting group still in place, which requires a linker far more acid-labile than the peptide's own protecting scheme. 2-Chlorotrityl chloride resin is the handle used for this in several of the syntheses below. Esterification proceeds in high yield with 0.6 equivalents of Fmoc-amino acid per mmol of resin in DCM or DCE in 25 min at room temperature, without by-product formation even for Fmoc-Asn and Fmoc-Gln, and cleavage with AcOH/TFE/DCM mixtures is quantitative within 15–60 min at room temperature while t-butyl type protecting groups remain unaffected [4]. The 2-chlorotrityl cations released during that cleavage do not attack the nucleophilic side chains of Trp, Met and Tyr [4], which matters because the fragment is about to spend hours in solution.
Concrete numbers from a three-fragment assembly of the C-terminal 21-residue sequence of HIV-1 protease: all three fragments were built manually on 2-chlorotrityl chloride resin in polypropylene syringes, and the two to be condensed in solution were cleaved with AcOH/TFE/DCM 1:1:8 at 5 mL per 100 mg of resin for 30 min, giving an octapeptide fragment in 67% yield at an estimated 92% HPLC purity and a nonapeptide fragment in 88% yield at 96% [5]. Trt protection of asparagine was found necessary in two of those fragments to prevent formation of the β-cyanoalanine [5].
Acid lability is not the only route. Under Boc/Bzl chemistry, where the side-chain groups are themselves acid-labile, a base-cleavable anchor does the same job: an N-[9-(hydroxymethyl)-2-fluorenyl]succinamic acid linker releases fully protected segments bearing a free α-carboxyl and requires no special equipment, at the price that there must be high compatibility between the protecting groups of the segment and an anchor cleaved by morpholine or piperidine in DMF [6].
Whether to protect the fragment's own C-terminus is a system-specific decision, and the literature does not point one way. A pseudo-Wang handle that leaves the C-terminus protected after cleavage with 2–4% trifluoroacetic acid was introduced specifically because it prevents polymerisation reactions in the convergent condensation of peptidic fragments, and it is traceless, being removed during final deprotection [7]. Against that, in the synthesis of the 48-residue Mdm2 RING finger domain both on-resin and solution fragment condensation were examined, the solution route with either a protected or a free C-terminus on the amino component, and the best results were obtained with the C-terminal carboxyl group left unprotected [14].
Epimerisation at the activated residue
Activating a fragment's C-terminal carboxyl opens a route to a 5(4H)-oxazolone, and that is where configuration is lost. The oxazolone is described as formed from an activated carboxyl component during the coupling [9], and the cleanest evidence that ordinary peptide segments take that route is the contrast with depsipeptides: peptides bearing a C-terminal depsipeptide unit couple without giving rise to epimerisation at the activated amino acid, attributed to the low tendency of the activated depsipeptide units, in contrast to the corresponding peptide segments, to form optically labile oxazolones [8].
What makes this worse in a convergent route than in stepwise assembly is exposure time. A fragment condensation onto a resin-bound amine can be slow at conventional dilution: in the HIV-1 protease work each condensation under conventional dilution was run for 18 h and then recoupled with fresh reagent for a further 18 h, and the Kaiser ninhydrin test was still positive afterwards [5]. Whatever epimer forms during that window is a diastereomer of the entire assembled product, not of a single residue's worth of resin.
The magnitude is not academic. In a silver-ion-activated thioester condensation between a 45-residue thioester segment and a 44-residue amino component, the epimer content of the 89-residue coupling product was reduced from 27% to 6% by moving the condensation to a less polar solvent [17]. That figure belongs to thioester activation of partially protected segments, not to carbodiimide activation of a fully protected fragment, and it should not be read as a general epimerisation rate for fragment condensation — but it sets the scale of what an unfavourable junction and an unfavourable solvent can do together.
The first line of defence is therefore to choose junctions where there is nothing to lose. Glycine is the obvious one, and it is used as an explicit design rule rather than a rule of thumb: in the HIV-1 protease assembly the three fragments were designed according to the positions of glycine to avoid racemisation at the C-terminus of a condensed fragment [5]. The second line is to place at the junction a residue whose activated form resists closing to the oxazolone, which is the rationale behind depsipeptide junctions; the same work also carries C-terminal pseudoproline units into segment coupling and compares them directly with depsipeptide units [8].
Neither trick is free. Base-assisted activation of depsipeptides with HBTU proceeds not only without loss of configuration but much faster than coupling of the corresponding all-amide segments; however, when the coupling of long depsipeptide segments proceeds slowly, β-elimination occurs at the activated depsipeptide unit, to an extent dependent on the presence of base in the system and on the type of solvent [8].
Reagents, additives and solvent
The suppression measures that work at this junction act on the same two levers: keep the oxazolone from forming, and keep the medium from racemising whatever oxazolone does form. Carbodiimide with a hydroxy-azole additive is the workhorse; DIC/HOBt is what was used for the on-resin condensations above [5]. Adding copper(II) chloride alongside HOBt in carbodiimide couplings was reported to eliminate racemisation in cases where a low level was seen with copper(II) chloride alone, and in a 4+1 segment condensation with both additives present no racemisation was detected and the yield was high enough; the benefit is ascribed to reduced 5(4H)-oxazolone formation together with suppression of racemisation in the oxazolone that remains [9]. That evidence comes from model couplings and one 4+1 condensation, so it does not license a claim that the combination is generally superior for long fragments.
Uronium reagents can win a head-to-head on a specific junction. In a 3+2 route to a nonapeptide run at multihundred-gram scale under Boc and benzyl ester protection, several coupling reagents were investigated in order to reduce the amount of epimerisation of the fragment coupling and TDBTU was identified as the condensation reagent of choice, the route delivering greater than 97% pure final product in 35% overall yield calculated on the di-Boc protected diaminosuberic acid [10]. For solution-phase assembly of much longer chains from roughly 10-residue protected segments, the reported combination is 1-ethyl-3-(3'-dimethylaminopropyl)carbodiimide with 3,4-dihydro-3-hydroxy-4-oxo-1,2,3-benzotriazine [11].
Solvent is a variable of the same rank as the reagent, not a detail. β-Elimination at an activated depsipeptide unit was completely suppressed by carbodiimide/HOBt activation in DCM, and in more polar media it was limited by substituting TMP for DIEA during HBTU activation or by using solvent mixtures such as DMSO/toluene for carbodiimide activation [8]. In the thioester system, epimerisation appeared during segment condensation in DMSO in the presence of a base and during thioester–thiol exchange, while thioester–thiol exchange in aqueous solution gave none [17].
Solubility as a practical limit
A fully protected 8–10-mer carries a large mass of tBu, Trt, Pbf or Bzl groups and behaves quite differently from the free peptide. The authors of the swelling-volume protocol name this directly: limitations of the approach include the solubility of activated fragments used and the steric demands of the protecting groups, which may critically influence the reaction yields [5]. Solution-phase convergent synthesis confronts the same problem head-on. The strategy that produced polypeptides of 123, 121, 136 and 238 residues assembles the entire molecule from fully protected segments in the size range of about 10 residues, each given a common Boc-peptide-OPac structure with Bzl-based side-chain protection, and depends on new types of solvent systems described for dissolving fully protected segments in which the segment condensation reactions can be carried out smoothly [11]. Segment length, protecting-group set and solvent are chosen together, not sequentially.
The other handle is composition. Distributing the hydrophobic residues and the bulky Arg(Pbf) across several fragments rather than concentrating them in one was the explicit basis of a successful four-fragment assembly [12]. Fragment boundaries therefore answer to at least three constraints at once: a glycine or otherwise oxazolone-resistant residue at each C-terminus, a tolerable count of difficult residues per fragment, and a protected fragment that will actually dissolve.
On-resin versus solution coupling
On-resin condensation lets excess reagent be washed away, but it puts the reaction at whatever concentration the vessel dictates, and that turns out to dominate. Dissolving the protected fragment and HOBt in the minimum, experimentally determined volume of DMA needed to swell the resin, adding DIC, and soaking a thoroughly dried resin in that solution gave a final fragment concentration of 30 mM at only 1.5 equivalents molar excess — roughly five times the concentration reached in a conventional volume at the same molar excess [5]. A single condensation in the swelling volume proceeded up to 93% and gave a negative Kaiser test after 18 h, against 40% overall for the same condensation repeated in a conventional volume; across two consecutive condensations the 21-residue product was obtained at 81% yield in a swelling volume compared with 21% under conventional conditions [5].
The on-resin route scales to multi-fragment assemblies. Hirudin variant 1, a 65-residue protein with three disulfide bonds, was assembled by sequential condensation of seven protected fragments onto the resin-bound 55–65 fragment on 2-chlorotrityl resin [13]. Solution coupling is the alternative when the amino component itself is best handled free, as in the Mdm2 RING finger case where solution condensation with an unprotected C-terminal carboxyl gave the best results and was judged well suited to large-scale synthesis [14]. A third arrangement splits the difference: prepare homogeneous protected segments rapidly on the solid phase, then couple them in solution [6].
Characterising the fragments before you condense them
Highly homogeneous segments are indispensable for a convergent strategy, whether the assembly runs on a resin or in solution [6]. That is a procedural requirement, not an aspiration: an impurity carried into a fragment is multiplied by every subsequent condensation and can no longer be separated from the target by a single residue's difference. Protected fragments were characterised by FAB-MS in the protected form — found 1944.5 against 1944.2 calculated for M+H+ for one octapeptide, 1088.8 and 1110.8 for M+H+ and M+Na+ for a nonapeptide — together with amino acid analysis after hydrolysis in 6 M HCl at 110 °C for 20 h and RP-HPLC purity at 222 nm, and the resin-bound C-terminal fragment was checked by cleaving and fully deprotecting a small sample [5].
Deletion tracking on the support is worth building in. End-capping the unreacted resin-bound amino groups after each fragment condensation with 2,4-dinitrofluorobenzene installs a substituted dinitroaniline chromophore on exactly the truncated chains, which are then read directly at 360 nm and separated, giving superior separation of the desired peptide [5]. In the conventional-volume condensation the capped truncation product was isolated at 29% against 40% for the target, so the label reported a real and large failure rather than a trace [5]. For the assembled chain, the reduced hirudin intermediate was purified to homogeneity by RP-HPLC before any disulfide work was attempted [13].
Why this is not native chemical ligation
Chemoselective ligation solves a different problem with a different constraint set. In native chemical ligation, chemoselective reaction of two unprotected peptide segments gives an initial thioester-linked species, and spontaneous rearrangement of that transient intermediate yields a full-length product with a native peptide bond at the ligation site [15]; in its classical form the two segments carry an N-terminal cysteine residue and a C-terminal thioester respectively, a pairing that thiol- and selenol-derived amino acid surrogates and ligation–desulfurisation chemistry have since widened [16]. Three operational differences follow. The segments are unprotected, so solubility is an ordinary aqueous peptide problem rather than a protected-fragment problem. The acyl donor arrives already activated as a thioester and the amide bond forms by rearrangement, so no coupling reagent is introduced at the junction. And the junction residue must be one the ligation chemistry supports rather than any residue taken from the sequence, which is the mirror image of the glycine-junction rule.
A thioester on a segment is not by itself a sign that chemoselective ligation is in use, and the distinction has consequences for stereochemistry. In peptide thioester condensation a C-terminal peptide thioester is selectively activated by silver ions and then condensed with an amino component, and the amino acid adjacent to the thioester is at risk of epimerisation [17]. In that system no substantial epimerisation of the C-terminal amino acid was detected during preparation of the thioesters by the Boc solid-phase method; it appeared during the thioester–thiol exchange reaction and during segment condensation in DMSO in the presence of a base [17]. Activation at the junction, not the presence of a thioester, is what puts the C-terminal stereocentre at risk.