Liquid Phase Peptide Synthesis via One‐Pot Nanostar Sieving (PEPSTAR)

Abstract Herein, a one‐pot liquid phase peptide synthesis featuring iterative addition of amino acids to a “nanostar” support, with organic solvent nanofiltration (OSN) for isolation of the growing peptide after each synthesis cycle is reported. A cycle consists of coupling, Fmoc removal, then sieving out of the reaction by‐products via nanofiltration in a reactor‐separator, or synthesizer apparatus where no phase or material transfers are required between cycles. The three‐armed and monodisperse nanostar facilitates both efficient nanofiltration and real‐time reaction monitoring of each process cycle. This enabled the synthesis of peptides more efficiently while retaining the full benefits of liquid phase synthesis. PEPSTAR was validated initially with the synthesis of enkephalin‐like model penta‐ and decapeptides, then octreotate amide and finally octreotate. The crude purities compared favorably to vendor produced samples from solid phase synthesis.


Introduction
Recent strategies to improve the pharmacokinetics and oral availability of peptide drugs have created arenaissance in peptide therapeutics.W ith more than 70 peptide drugs approved for clinical use and many in the development pipeline,g lobal peptide therapeutic demand is rising. [1] This trend has created an imperative for continuing innovation in peptide manufacturing technology in anticipation of growing future demand for production of peptide therapeutics.
When the exact sequence of monomers in ap olymer is crucial, as in peptide synthesis,c hemical approaches utilize iterative synthesis with protecting group chemistry to ensure high sequence precision, with separation of unreacted monomers from the growing polymer at each cycle being the crucial step.S olid phase attachment solves this key separation problem for peptide synthesis,e nabling excess amino acids to be washed away and separated from the support-bound growing peptide.S olid phase peptide synthesis (SPPS) provides for facile manipulation of intermediates retained in the solid support, and facilitates process automation and rapid synthesis in asingle reactor. [2] Hence SPPS is widely regarded as the gold standard for peptide synthesis.H owever,t he intraparticle reactions also bring penalties.T hese include diffusional limitations in the solid supports leading to incomplete couplings and deprotections as well as the use of reagent in large excesses to drive the diffusion and reaction in the solid phase. [3] Liquid phase peptide synthesis (LPPS) could potentially accomplish higher crude purity,l ower reagent consumption, and greater ease of scaling than SPPS. However,t hese advantages have not so far overtaken SPPS because the key separation of intermediates from reaction byproducts is usually achieved by precipitation or extraction. These relatively time consuming operations must often be optimized from one cycle to the next, and from one target to another, as the physical properties of growing peptide intermediates vary unpredictably,making the overall process slow and tedious. [4] Thefirst LPPS strategy to address these deficits attempted to standardize the process by anchoring all intermediates permanently in an organic phase,r emoving by-products by extraction or precipitation, and with sequential couplings and deprotections requiring just one isolation step at the end of synthesis. [5] To further control the solubility of growing peptides,r ecent developments in hydrophobic anchors for LPPS have increased the practicality of extraction and precipitation approaches. [5a-e] These works demonstrated the expected fast coupling kinetics,a nd utility of reaction monitoring to ensure high purity,b ut precipitation and extraction methods have still not been widely adopted. Multi-step work-ups,s uch as washing with several aqueous phases,and material transfer between apparatus lack process continuity ( Figure S1). Generally,isolation via phase transfer, either liquid-to-liquid or liquid-to-solid, is challenging for scale-up,thus prolonging process development lead-time and impeding automation. [6] One pot (i.e.telescoped) LPPS with no phase or material transfer between synthesis cycles would be ideal for peptide manufacturing.A nL PPS approach built around membranebased separation, or diafiltration, might provide this.B ayer et al. were the first to describe this approach, but their process was impractical due to the lack of solvent stable membranes; solvent exchange into water was required prior to purification using ad ialysis membrane. [7] Membrane-based LPPS is greatly facilitated if the separations can be conducted in the same organic solvent the reactions are performed in, and recent advances in organic solvent nanofiltration (OSN) have brought this approach within reach. [8] Membrane-enhanced peptide synthesis (MEPS), was introduced by So et al. employing al inear 5kDa methoxy polyethylene glycol (mPEG) anchor to maximize peptide retention. [9] After each coupling then deprotection step, reaction by-products and excess reagents were permeated through as olvent stable membrane,w hile the growing peptide-mPEG conjugate was retained by the membrane within ac ombined reactor-separator.A lthough the MEPS concept was validated through the synthesis of apentapeptide, the process was compromised by relatively low mass efficiency and low anchor loading capacity ( % 0.2 mmol g À1 ). Excessively large linear mPEG was required to minimize yield loss through the membrane attributed to the ability of PEG to adopt an extended, instead of aglobular random coil, conformation that was membrane permeable. [10] Recently, MEPS was revisited using al arge permanently globular anchor ( % 6-8 kDa) to improve the separation. [6b] Despite having multiple conjugation sites at the ends of four or five polymer arms,t he loading capacity of the anchor was low ( % 0.6 mmol g À1 ). Furthermore,t he polydispersity of the anchor resulted in ab road molecular weight distribution and diffuse range of properties,making quantitative reaction monitoring technically challenging,n egating one potential major advantage of LPPS.T ot he best of our knowledge,n o unimolecular anchor has been reported for membrane-based LPPS. [11] We believe that these limitations have deterred the wider adoption and development of MEPS.
In this report, inspired by recent successes in membranebased liquid phase synthesis of sequence-defined polymers, [12] we introduce liquid phase peptide synthesis via one-pot nanostar-sieving (PEPSTAR,F igure 1). Several innovations are incorporated to address the shortcomings identified in earlier membrane-based LPPS protocols:1 )Acompact, monodisperse "nanostar" hub was designed;2 )r eal-time reaction monitoring by UHPLC-MS was undertaken;a nd 3) an efficient one-pot process operation was developed. This new LPPS concept is promising for development into af ully automated peptide manufacturing technology.

Results and Discussion
Nanostar design and synthesis.Acompact, easily synthesized hub was needed that would nevertheless increase the molecular size of peptide intermediates for efficient nanofiltration. At hree-armed, star-shaped molecule,o r" nanostar", was selected to provide the desired difference in molecular weight between the growing peptide and the reaction by-products,w hile at the same time minimizing the number of chain extension intermediates needing to be resolved during UHPLC-MS reaction monitoring. Thet hree growing peptides are linked by an aromatic hub,w hich also acts as an additional UV chromophore,b yf lexible,m onodisperse octaethylene glycol (octagol, HO-Eg 8 -OH) spacers, to limit steric interaction between adjacent peptides [13] (Figure 1). Compared to earlier MEPS strategies,t his design increases the loading capacity of the anchor with three peptide attachment sites;increases the mass efficiency due to the low molecular weight of the anchor;a nd enhances the separation due to the at least 3-fold mass difference between Figure 1. Key features of PEPSTAR:A ctivatedF moc-amino acid building blocks couple to chain termini of H-peptide-nanostar;p iperidine quenches excess amino acid and removes Fmoc;d iafiltration washesa ll reaction by-products through amembrane;p urified H-peptide-nanostar is retained in the synthesizer,r eady to repeat the cycle. Every process is analyzed in real time by UHPLC-MS. nanostar and excess amino acids (AA). Furthermore,t he unimolecular,f ully defined composition and UV chromophore are commensurate with reaction monitoring by liquid chromatography and mass spectrometry.
Synthetic strategy.The widely used Fmoc-peptide chemistry was adopted for PEPSTAR. [16] In classical LPPS Fmocpeptide synthesis,four steps are typically required:Coupling, isolation, deprotection and isolation again. Aiming for aonepot synthesis strategy with sequential couplings and Fmoc removal, we elected to omit the post-coupling purification step by quenching excess amino acid (AA) active ester.T his has the advantages of reducing both solvent consumption and total cycle time.Asacrificial species is required to react with the active ester to prevent uncontrolled peptide chain extension after Fmoc removal. Initially,i nt he reaction of H-Thr(tBu)-O-Wang-nanostar 8 with Fmoc-Asn(Trt)-OH, aniline was tested for this purpose because of its low basicity and we believed it would give aU V-active by-product. Unfortunately,t he reaction of aniline with active esters was Scheme 1. Synthesis of nanostars and chain extension cycle:a)T ris-hydroxy terminated octagol-nanostar 1 is converted to Wang-type, 4,and Rink, 6,n anostar;t he Wang anchor is loaded by slow esterification with the first AA before entering the synthesizer.b)Using the Fmoc strategy, peptide-nanostars are grown in athree-step cycle of coupling, Fmoc removal, and diafiltration, then removed from the synthesizer for global deprotection. too slow to afford effective quenching.H owever, despite dispensing with the active ester quench after deprotection of Fmoc-Asn(Trt)-Thr(tBu)-O-Wang-nanostar 9 with piperidine,n od ouble coupling contaminants could be detected in the product H-Asn(Trt)-Thr(tBu)-O-Wang-nanostar 10 (Figure 2). We observed that remaining active ester was converted instantaneously by excess piperidine into H-Asn(Trt)-Pip 11 (Scheme 1b). [17] Thereaction of excess AA with piperidine eliminated the need for both an additional sacrificial species and as eparate diafiltration post-coupling. Thus,asynthesis cycle requiring only one diafiltration step was devised, as illustrated in Scheme 1. After charging the synthesizer with nanostar, the chain extension cycle was initiated:T he first Fmoc-amino acid and condensing agent were injected into the synthesizer; after UHPLC-MS confirmed the coupling was complete, Fmoc removal was initiated with piperidine;f inally, after UHPLC-MS confirmed complete deprotection, diafiltration commenced to remove the H-AA-Pip 11 derived from excess AA, the DBF-Pip adduct 12,a nd other reaction byproducts,w hile the H-AA-nanostar was rejected by the OSN membrane and remained in the synthesizer.T his process was repeated for every new amino acid in the sequence up to H-(AA) n -nanostar,and resembles SPPS in its simplicity.
Nanostar coupling and Fmoc removal kinetics. Chain extending apeptide-nanostar proceeds to completion via two intermediate species.T aking the formation of Fmoc-dipeptide-nanostar 9 as an example (Figure 2a give asecond species with two arms coupled (2-arm, 14), but with one amino-terminated arm still free.F inally,t he 2-arm intermediate reacts on the remaining amino group to give the desired Fmoc-dipeptide-nanostar 9.E ach of these species is chemically distinct and chromatographically resolvable.B efore examining complex peptide-nanostars,t he variation in reaction rates for forming ar ange of dipeptides,a nd their subsequent Fmoc removal, was examined in various solvents (see SI section 4). Suspecting that as ingle solvent could not Figure 2. Variation in reaction rates for ad imer-nanostar:a )The intermediates of peptide-nanostar chain extension.b)The variation in the average rates of coupling to form Fmoc-AA-Phe-OMe and of Fmoc removal with solvent polarity;tested with AA = Val, Leu, Glu(Trt), Asp(tBu), Trp(Boc), Arg(Pbf), Ala. As ageneral trend, solubility of H-dipeptides increases with solvent polarity.c-e) Variation in concentration with time of 1-arm 13 and 2-arm 14 intermediates, and Fmoc-dipeptide-nanostar 9,respectively,from different substrate concentrations;red 1wt%,blue 2wt% and yellow 5wt% starting concentration of H-Thr-nanostar 8.N Bt he drop in relative absorbancefor 2wt% after 125 mins in (e) was most likely caused by sample preparatione rrors, high dilution was necessary in order to quench the reaction prior to the UHPLC-MS analysis.
optimally fulfil all the required roles,binary mixtures of THF and NMP in different proportions were also considered. [14] It was found that there is as trong correlation between solvent polarity and reaction rate;l ow polarity leading to fast coupling and slow deprotection, and vice versa in high polarity solvents (Figure 2b). It was also noted that the urea by-product (DIU) of AA activation with DIC and HOBt, and dipeptides after Fmoc removal, had poor solubility in low polarity solvents. [18] Therefore,aTHF-NMP 35:65 v/v mixture was selected to provide ab alance between reasonably fast reaction kinetics,f or both coupling and Fmoc removal, and solubility,and was used for astudy of the kinetics of peptidenanostar coupling and Fmoc removal.
Kinetic runs at three different concentrations of [H-Thr(tBu)-O-Wang] 3 -nanostar 8 (1, 2a nd 5wt%)w ere performed in ac arousel reactor with THF-NMP 35:65 v/v solvent (See SI section 5). Coupling reactions containing 1wt% 8 and 5equiv Fmoc-Asn(Trt)-OH (i.e.1 .7 equiv per amino terminated arm) went to completion smoothly:t he 1arm intermediate formed almost instantaneously (Figure 2c), but the time needed for the 2-arm species to drop below both UHPLC and MS detection was over 4hours (Figure 2d). The coupling was also carried out at higher concentrations of nanostar 8 where the time for complete reaction to 3-arm Fmoc-dipeptide-nanostar 9,that is,for 2-arm intermediate 14 to disappear,w as reduced markedly (Figure 2e): at 2a nd 5wt% the coupling was finished within one hour and 2minutes,respectively.Upon addition of 10 v/v% piperidine, Fmoc removal was much faster than coupling,being complete < 10 minutes (Table S3). These highlight the importance of nanostar concentration to reaction kinetics,d efining the lower boundaries for nanostar concentration.
Membrane selection and synthesizer design. TheO SN membrane is at the heart of PEPSTAR technology.T hree chemically robust polymeric membranes developed in our laboratory,c apable of permeating larger solutes,w ere screened for OSN:Apolyethyleneimine (PEI) asymmetric membrane,c ross-linked with terephthalic chloride (TPC), and coated with Jeffamine M-2005 (a polyether monoamine), denoted PEI_2005;t wo different batches of polybenzimidazole (PBI) asymmetric membrane,c ross-linked with a,a'-p-dibromoxylene (DBX), and modified with apolymer brush Jeffamine M-2005, denoted PBI_2005(1) and PBI_2005(2). [19] These surface-modified membranes offer anti-fouling properties to reduce the undesirable deposition of solutes on the surface. [20] Thek ey difference between PBI_2005(1) and PBI_2005(2) was that the latter was stored in acetonitrile for longer period prior to cross-linking, allowing the freshly cast PBI film to tighten to varying extents.Membrane screening was conducted in THF which is ag ood solvent for peptide intermediates,a nd the rejections were calculated based on Equation (S3) (Figure 3a,s ee also Figure 3. Synthesizer design and separation performance:a)Rejection of peptide-nanostars 8 and 10,a nd reaction by-products 11 and 12,by candidate membranes. b) Modelling the retention and purity of dipeptide-nanostar 10 in single stage or two-stage membrane separators containing PBI_2005(1);t he early dip in the two-stage yield curve is aresult of redistribution of asmall proportiono f10 from stage 1t ostage 2. Diavolume is the ratio of cumulative volume of wash solvent introducedt othe synthesizer at any given time of diafiltration per volume of stage 1( DV = Wash solvent volume /stage 1volume), 1DV= 200 mL. c) Schematic of synthesizer layout, with picture of the synthesizer setup shown in Figure S8. d) Purificationo fH-dipeptide-nanostar 10 in the synthesizer,f rom the largest by-products H-Asn(Trt)-Pip 11 and DBF-Pip 12;tothe 2-arm by-productfrom diketopeparazine (DKP) can be detected, although this does not affect final peptide purity. SI section 6). As expected from their MWs,H -Asn(Trt)-Pip 11 always had ahigher rejection than DBF-Pip 12,therefore the critical separation to achieve was selective permeation of residual building block while retaining the nanostar. The efficiency for separating molecule Af rom Bi se xpressed by the separation factor, b A/B [Eq. (S4)] the higher it is the better the separation. PBI_2005(1) gave the highest separation factor b 10/11 for purifying H-dipeptide-nanostar 10 from byproduct 11 (9.2) compared to both PBI_2005(2) (4.2) and PEI_2005 (4.7);t he separation factor for smaller H-AAnanostar 8 and by-product 11, b 8/11 ,w as again highest for PBI_2005(1) (6.1). Although PBI_2005(2) gave as imilar peptide-nanostar rejection to PBI_2005(1), the separation factor is much lower due to its simultaneously higher rejection of by-product. Hence,P BI_2005(1) was selected for further studies. Figure 3ashows that the rejection rises with MW,but only approaches 100 %r ejection slowly.T hus,e ven though PBI_2005(1) gave the highest separation factor,asignificant loss of yield is inevitable during diafiltration, particularly during the early cycles of synthesis when the MW of the peptide-nanostar is relatively low.F or this reason, membrane "cascade" systems were conceived that used membranes linked in series to overcome the performance limits of asingle membrane by recycling partial flows back to previous stages. Building on this concept, ap ractical two-stage separator, requiring as ingle HPLC pump to pressurize the equipment, and using pressure relief valves to control the pressure in each stage,w as implemented for the first time by Kim et al. [21] Modelling,b ased on the measured rejection value for PBI_2005(1), predicted that if apurity of 90 %was required, then at least 40 %o fd ipeptide-nanostar 10 would be lost to permeation using as ingle stage diafiltration system (Figure 3b). However,i fat wo-stage separation system was adopted using the same membranes,y ield loss improved to about 10 %w hile achieving the same purity.O ur synthesizer is based upon this two-stage design, which is ac ompromise between achievable yield, complexity,a nd the time required for diafiltration (which rises with the number of stages).
Thek ey difference between the synthesizer described here and earlier membrane separation apparatus [12] is that the chemistry to assemble the heteropolymer,i nt his case ap eptide,o ccurs inside the same equipment as is used to undertake purification of the crude product from reagent and by-products.T he principal elements of the synthesizer are ( Figure 3c): Thef eed-tank, into which fresh solvent is passively added during OSN from ar eservoir at the same rate as the waste stream from OSN drains to acollection tank (providing constant volume diafiltration);t he filling of the waste tank is monitored electronically to enable automatic shut-down. An HPLC pump takes liquid from the feed tank to pressurize the stage 1c irculation loop containing two membrane cells.The pressure is regulated by apressure relief valve that returns liquid from the stage 1loop back to the feed tank. Thepermeate from stage 1enters the stage 2circulation loop where the pressure is again controlled by ap ressure relief valve.T he stage 2membrane concentrates the peptidenanostar permeating from stage 1a nd ap artial recycle flow returns to stage 1, increasing yield. Pumps in both stages circulate at 90 Lh À1 ,t om inimize concentration polarization at the membrane surface.
Thesynthesizer was validated with asynthesis run to form dipeptide-nanostar 10.The membrane cells of the synthesizer were fitted with disks of PBI_2005(1), and stage 1was charged with H-Thr(tBu)-O-Wang-nanostar 8 in THF.Onaddition of DIC,H OBt and Fmoc-Asn(Trt)-OH to the reactor,t he coupling proceeded smoothly to completion, as did the subsequent Fmoc removal upon addition of 10 %v/v piperidine.O nce the chemistry was complete,t he valve to stage 2 was opened commencing diafiltration (see SI section 7). After 10 DV of solvent had permeated H-Asn(Trt)-Thr(tBu)-O-Wang-nanostar 10 was completely purified with by-products below the detection limit (Figure 3d). To test the general applicability of the method, we extended the synthesizer validation to different H-AA-Pip (see Figure S9). Results showed that 10 DV were sufficient to remove most H-AA-Pip.During optimization and several repeats of this protocol the membrane exhibited constant performance,a nd later visual inspection confirmed no signs of chemical or physical degradation.
Proof of concept. With the kinetics optimized, it was necessary to define an initial target for preparing longer peptides that would assist in validating the operation of the new synthesizer. Although the HO-Wang-nanostar 4 is essential for preparing peptides with C-terminal carboxylic acids,aswith its counterparts among SPPS supports,ati= 2it suffers from low to moderate peptide cleavage from the support during extension, due to base catalyzed cyclization to form the diketopiperazine (DKP). [22] Although for most sequences the yield loss is low ( % 10 %a sd etermined on chromatograms for example,F igure 3d), it introduces additional signals into the subsequent UHPLC chromatograms, corresponding to (peptide-O-Wang) 2 -(HO-Wang)-nanostar, that will confound the analysis of an ew system. Fort his reason, we changed to the H-Rink-nanostar anchor,used for preparing C-terminal amides,b ecause it does not produce analogous DKP by-products.T ov alidate the process of repeated chain extension cycles during the development phase of PEPSTAR, an enkephalin-like model peptide sequence was designed ( Figure 4a). Theg lycine residues of the native Leu-enkephalin (H-Tyr-Gly-Gly-Phe-Leu-NH 2 ) sequence were replaced by serine to make (Ser2,Ser3)-Leuenkephalin pentapeptide (H-Tyr-Ser-Ser-Phe-Leu-NH 2 ), because H-Gly-Pip displayed lower solubility in THF-NMP 35:65 than other H-AA-Pip;this obstacle may be avoided by nanofiltering Fmoc-Gly-OH prior to deprotection via an additional diafiltration. Stage 1o ft he synthesizer (total % 200 mL) was charged with unloaded H-Rink-nanostar 6 ( % 1.7 wt %) dissolved in THF-NMP 35:65. Thes ame protocols were then used to run the synthesizer,e xcept that the temperature was maintained at 35 8 8Cf or all reactions via aheater feedback control loop.Fmoc-Leu-OH (5 equiv,that is,1 .7 equiv per amino terminated arm) in the minimum amount of solvent, preactivated for 1minute with DIC and HOBt, was injected into stage 1ofthe synthesizer via the feed tank. Once UHPLC-MS confirmed completion of the coupling, piperidine (20 mL, 10 v/v%) was added to the feed tank to initiate Fmoc removal and quenching of the active ester; the piperidine was added slowly to further ensure that all the AA was trapped before Fmoc removal commenced. As before,d iafiltration was initiated after reaction completion, but now in THF-NMP.Once the H-Leu-Rink nanostar (i = 1) had been purified, the synthesis cycle was repeated four more times to obtain H-pentapeptide-nanostar 16 (i = 5, Figure 4a).
All reactions and OSN were monitored in real-time with UHPLC-MS.A sb efore,t he 1-arm and 2-arm intermediates, as well as the 3-arm desired peptide-nanostar,a re well resolved on the chromatogram (see Figure 4b for i = 1t o2 ), permitting easy and sensitive reaction monitoring,a nd the same is true for subsequent Fmoc removal;the peak identities were corroborated by mass spectrometric characterization ( Figure S10). Most couplings went to completion within 1hour,b ut to minimize traces of deletion sequences the reactions were allowed to run for an additional hour. Much faster Fmoc deprotection was largely complete within 10 mins,but was allowed to run for 30 mins,again to minimize chain length errors.M onitoring of purification showed that (e.g. Figure 3d)a fter 8 diavolumes (1 DV % 200 mL) had permeated all large MW reaction by-products were removed, including H-AA-Pip and DBF-Pip.H owever,p urifications were extended to 10 DV to reduce residual piperidine to below 0.1 v/v% (as determined gas chromatography);t his is essential to prevent decomposition of fresh Fmoc-AA-OH added at the start of the next cycle.A st he peptide-nanostar grew in length, its rejection also increased and reached virtually 100 %f rom i = 8o nwards (Figure 4c). Diafiltration in our small synthesizer typically required 15 hours,due to the relatively low membrane area/volume ratio of flat sheet cells. Thus,a utomatic shutdown ensured that one cycle could be undertaken reliably every 24 hours by running most of the OSN purification overnight. Although OSN was slower in THF-NMP than THF alone,d ue to the higher viscosity of NMP,the separation was very efficient, possibly enhanced by the bulky Rink amide moiety.
Global deprotection of H-pentapeptide-nanostar 16 was undertaken in TFA-TIS-H 2 O38:1:1(Scheme 1b), after which the crude pentapeptide H-Tyr-Ser-Ser-Phe-Leu-NH 2 17 ex- hibited 94 %p urity (Figure 4d). This level of purity,u sing only 1.7 equiv Fmoc-AA-OH per arm every cycle,i sc omparable with SPPS vendor using 3.0 equiv AA to obtain 97 % crude purity.Encouraged by this success,the H-pentapeptidenanostar 16 was further elongated to ad ecapeptide by repeating the model (Ser2,Ser3)-Leu-enkephalin sequence ( Figure 4a). Thes ame PEPSTAR protocols were applied, with UHPLC-MS monitoring of reactions and OSN.A ti= 9 the nonapeptide-nanostar UHPLC peak shape became broader and mis-shaped, with lower MS sensitivity (although reaction progress could still be determined by changes in retention time), probably due to poor solubility of the hydrophobic peptide-nanostar in the part-aqueous mobile phase ( Figure S10). This time global deprotection of decapeptide-nanostar 18,afforded crude decapeptide H-(Tyr-Ser-Ser-Phe-Leu-) 2 -NH 2 19 with 84 %p urity from 1.7 equiv, considerably improving on the crude purity of 77 %f rom SPPS vendor, again 3.0 equiv AA per cycle (Figure 4e). The decapeptide yield was 75 %. Although % 5% of the loss can be accounted for due to frequent sampling from the synthesizer,t he majority was lost in the first five cycles, particularly the first, when the peptide-nanostar rejection was relatively low.
To investigate the effects of AA equivalents on SPPS,the excess of AA per cycle was reduced from 3.0 equiv to 1.7 equiv,s imilar to that used in PEPSTAR protocols. Although the pentapeptide 17 purity was unaffected by the reduced excess of AA, the crude purity of SPPS decapeptide 19 plummeted from 77 %t o35 %. Thei mpurities were identified to be deletion sequences,presumably caused by onresin peptide aggregation or conformational effects during the elongation from i = 5t o1 0w hich led to incomplete coupling or Fmoc removal ( Figure S12). However,t he authors emphasize that the solid phase synthesis with reduced excess of AA was not optimized to achieve the highest purities,b ut to mirror PEPSTAR quantities of reagents (see SI section 11).
To demonstrate the reliability of PEPSTAR, the model penta-and decapeptide syntheses were reproduced giving 17 and 19 of essentially the same quality and yield (both + /-1%), (Figure 4d,e(ii). Tw oa dditional runs were performed, but with the initial H-Rink-nanostar 6 at 1a nd 3wt%,t o evaluate the effect of concentration on the performance of the synthesizer.U nder identical conditions to those above,t he 1wt% Rink anchor solution resulted in lower purities of penta-and decapeptides 17 and 19 at values of 87 %and 60 %, respectively;t his is expected from slower kinetics as ar esult of higher dilution. At 3wt% initial nanostar concentration, the kinetics were fast and clean up to the pentapeptide,but at the Fmoc-octapeptide-nanostar, where the supported peptide concentration was % 9wt%,p lus af urther 3wt% reagents, gelation of the reaction solution occurred. As expected, any LPPS process has as olubility ceiling that varies from one target peptide to the next, but should be amenable to further optimization of solvents,t emperature and protecting groups. Fort he reported model peptide,t he optimum starting concentration for H-Rink-nanostar 6 is around 2wt%, corresponding to 7wt% of final H-decapeptide-nanostar 18.
Octreotate. After validation of the PEPSTAR synthesis system, we selected linear octreotate amide as amore typical peptide target. Starting with an H-Rink-nanostar 6 concentration of around 2wt%,t he octapeptide sequence was synthesized using the same protocols as above in THF-NMP 35:65. Notably,t he Fmoc-Cys(Acm)-OH building block was selected, aiming for Cys to withstand the acidic global deprotection with the Acm intact. All reactions and diafiltrations went smoothly with no solubility issues.T he fully protected peptide-nanostar 20 was obtained in 80 %yield, and after global deprotection the crude purity of linear octreotate amide 21 was 90 % (Figure 5c). During the synthesis towards peptide-nanostar 20,t he degree of epimerization was also investigated at the epimerization-prone Cys residue.T he global deprotection of H-Thr(tBu)-Cys(Acm)-Thr(tBu)-Rink-nanostar (i = 3) showed < 0.1 %o fe pimerization from L-to D-Cys ( Figure S14).
To further explore the capability of PEPSTAR, the synthesis was recapitulated with H-Thr(tBu)-O-Wang-nanostar 8 to produce linear octreotate with C-terminal carboxylic acid. To minimize the losses due to base-catalyzed DKP formation, extension cycle and subsequent Fmoc removal from i = 1t o2w ere conducted in just the feed tank, setting the nanostar and piperidine concentrations at ca. 10 wt %and 10 v/v%, respectively.Upon completion, the reaction mixture was immediately diluted into stage 1, lowering the nanostar and piperidine concentrations to 2wt% and 2v/v%. Diafiltration was then initiated, reducing the piperidine concentration in stage 1, thus minimizing the time that piperidine would have ah igh enough concentration to catalyze loss of dipeptide to DKP.E ven so,1 0% (H-dipeptide-O-Wang) 3nanostar was still converted to (H-dipeptide-O-Wang) 2 -(HO-Wang)-nanostar via DKP.F rom i = 2o nwards,t he standard protocol was applied. All reactions and diafiltrations went smoothly to obtain fully protected peptide-nanostar 22 in 71 %y ield. After global deprotection the purity of linear octreotate was 70 %, 88 %i fy ou include 18 %c yclized octreotate from loss of Acm, which is also comparable to SPPS (Figure 5d,s ee also Figure S16). Thef ast reaction kinetic facilitated in liquid phase was suspected to cause Acm removal and promote cyclization. Hence,f urther optimizations are required for the global deprotection protocol.
With successful completion of the proof of concept synthesis of octreotate amide and acid, it is pertinent to ask how the environmental and economic credentials of aputative commercial PEPSTAR system compare at the first pass with MEPS and its principal competitor SPPS (Figure 5e). Process Mass Intensity (PMI) is used to express solvent efficiency, which measures the total material used by mass per unit mass of peptide (kg kg À1 ). [23] ThePMI evaluation shows that, for the three methods under assessment, the solvent consumption contributed to nearly all the total waste generated with only as mall fraction contributed by hubs (or resins) and other reagents (see Table S6). TheP MI is drastically lower for PEPSTAR than for MEPS,a nd only slightly higher than for SPPS.T ocompare production costs,wesummed up the costs of materials for PEPSTAR and MEPS to produce 1mole of linear octreotate amide based on current supplier prices (Sigma-Aldrich). Using the same material prices,w et hen estimated the cost for the vendor of SPPS to produce similar quantity of product following the standard protocol (see SI section 11). Thecost of materials for SPPS are the highest due to the large excess of AA (3 equiv) required to achieve the specified purity (see Table S7). Furthermore,w ea nticipate that the PMI and cost of PEPSTAR will fall below the benchmark set by SPPS once downstream purification steps are considered. Theh igher crude purity offered by liquid phase synthesis in turn requires less demanding chromatographic purification, which is often am ajor contributor to waste generation. [24] Conclusion In this report, we have demonstrated liquid phase peptide synthesis via one-pot nanostar sieving (PEPSTAR), acontinuous process to synthesize high purity peptides,without phase or material transfers,s uitable for automation. This advance depended on the development of highly stable,p olymeric organic solvent nanofiltration (OSN) membranes that enabled reaction by-products to be efficiently "sieved" out from the crude reaction mixture to fully purify the growing peptide by diafiltration. Equally,t he design of at hree-armed starshaped monodisperse synthesis support, or nanostar anchor, for growing peptides facilitated efficient membrane separation and real time monitoring,o fb oth the synthesis and purification phases of PEPSTAR. This one-pot synthesis strategy lowered the PMI three-fold from its LPPS predecessor, membrane-enhanced peptide synthesis (MEPS), and is close to that of SPPS.T he cost of materials is estimated to be half of SPPSs. This augurs well for manufacturing scale-up of PEPSTAR which, because of its entirely liquid phase nature,i sf ar less scale-limited than SPPS,i se xpected to reduce the lengthy lead-times for process development, and the PMI and production cost are anticipated to fall further at larger scales.T his technology is highly flexible in terms of solvent and nanostar support choice.I nf uture,f ull process automation and further innovations on nanostar support and solvent system will realize efficient synthesis of large peptides or even proteins on PEPSTAR.