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

Article abstract of DOI:10.1002/anie.202014445

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.

Full text of DOI:10.1002/anie.202014445

Angewandte
Research Articles
Chemie
How to cite: Angew. Chem. Int. Ed. 2021, 60, 7786–7795
International Edition:
German Edition:
doi.org/10.1002/anie.202014445
doi.org/10.1002/ange.202014445
Peptide Synthesis
Liquid Phase Peptide Synthesis via One-Pot Nanostar Sieving
(PEPSTAR)
Jet Yeo, Ludmila Peeva, Seoyeon Chung, Piers Gaffney, Daeok Kim, Carla Luciani,
Sergey Tsukanov, Kevin Seibert, Michael Kopach, Fernando Albericio, and Andrew Livingston*
Abstract: Herein, a one-pot liquid phase peptide synthesis
step. Solid phase attachment solves this key separation
featuring iterative addition of amino acids to a “nanostar” problem for peptide synthesis, enabling excess amino acids
support, with organic solvent nanofiltration (OSN) for iso-
lation 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.
to be washed away and separated from the support-bound
growing peptide. Solid phase peptide synthesis (SPPS)
provides for facile manipulation of intermediates retained in
the solid support, and facilitates process automation and rapid
synthesis in a single reactor.[2] Hence SPPS is widely regarded
as the gold standard for peptide synthesis. However, the
intraparticle reactions also bring penalties. These 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, lower
reagent consumption, and greater ease of scaling than SPPS.
However, these advantages have not so far overtaken SPPS
because the key separation of intermediates from reaction by-
products 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]
The first LPPS strategy to address these deficits attempted
to standardize the process by anchoring all intermediates
permanently in an organic phase, removing 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, recent 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, and utility of reaction
Introduction
Recent strategies to improve the pharmacokinetics and
oral availability of peptide drugs have created a renaissance in
peptide therapeutics. With more than 70 peptide drugs
approved for clinical use and many in the development
pipeline, global 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 a polymer is
crucial, as in peptide synthesis, chemical approaches utilize
iterative synthesis with protecting group chemistry to ensure
high sequence precision, with separation of unreacted mono-
mers from the growing polymer at each cycle being the crucial
[*] J. Yeo, Dr. L. Peeva, S. Chung, Dr. P. Gaffney, Dr. D. Kim,
Prof. A. Livingston
and
School of Chemistry and Physics, University of KwaZulu-Natal
University Road, Westville Campus, Durban 4001 (South Africa)
Barrer Centre, Department of Chemical Engineering, Imperial College
London, Exhibition Road, London SW7 2AZ (UK)
E-mail: a.livingston@imperial.ac.uk
Prof. A. Livingston
School of Engineering and Materials Science, Queen Mary University
of London, Mile End Rd, Bethnal Green, London E1 4NS (UK)
Dr. C. Luciani
Eli Lilly and Company, RNA-Therapeutics
450 Kendal St, Cambridge, MA 02142 (USA)
Supporting information and the ORCID identification number(s) for
the author(s) of this article can be found under https://doi.org/10.
1002/anie.202014445.
Dr. S. Tsukanov, Dr. K. Seibert, Dr. M. Kopach
Eli Lilly and Company, Synthetic Molecule Design & Development
1200 W Morris St, Indianapolis, IN 46221 (USA)
ꢀ 2021 The Authors. Angewandte Chemie International Edition
published by Wiley-VCH GmbH. This is an open access article under
the terms of the Creative Commons Attribution License, which
permits use, distribution and reproduction in any medium, provided
the original work is properly cited.
Prof. F. Albericio
Networking-Centre on Bioengineering, Biomaterials and Nanome-
dicine, Department of Organic Chemistry, University of Barcelona
Marti i Franques 1–11, 08028 Barcelona (Spain)
7786
ꢀ 2021 The Authors. Angewandte Chemie International Edition published by Wiley-VCH GmbH
Angew. Chem. Int. Ed. 2021, 60, 7786 –7795

Angewandte
Research Articles
Chemie
monitoring to ensure high purity, but precipitation and
anchor resulted in a broad molecular weight distribution
extraction methods have still not been widely adopted.
Multi-step work-ups, such 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. An LPPS approach built around membrane-
based separation, or diafiltration, might provide this. Bayer
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 a dialysis 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]
and diffuse range of properties, making quantitative reaction
monitoring technically challenging, negating one potential
major advantage of LPPS. To the best of our knowledge, no
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 membrane-
based liquid phase synthesis of sequence-defined polymers,[12]
we introduce liquid phase peptide synthesis via one-pot
nanostar-sieving (PEPSTAR, Figure 1). Several innovations
are incorporated to address the shortcomings identified in
earlier membrane-based LPPS protocols: 1) A compact,
monodisperse “nanostar” hub was designed; 2) real-time
reaction monitoring by UHPLC-MS was undertaken; and 3)
an efficient one-pot process operation was developed. This
new LPPS concept is promising for development into a fully
automated peptide manufacturing technology.
Membrane-enhanced peptide synthesis (MEPS), was Results and Discussion
introduced by So et al. employing a linear 5 kDa 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 a solvent stable membrane, while the growing
peptide-mPEG conjugate was retained by the membrane
within a combined reactor-separator. Although the MEPS
concept was validated through the synthesis of a pentapeptide,
the process was compromised by relatively low mass effi-
ciency and low anchor loading capacity ( ꢀ 0.2 mmolgÀ1).
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 a globular random coil,
conformation that was membrane permeable.[10] Recently,
MEPS was revisited using a large 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, the loading capacity of the anchor was low
( ꢀ 0.6 mmolgÀ1). Furthermore, the polydispersity of the
Nanostar design and synthesis. A compact, easily synthe-
sized hub was needed that would nevertheless increase the
molecular size of peptide intermediates for efficient nano-
filtration. A three-armed, star-shaped molecule, or “nano-
star”, was selected to provide the desired difference in
molecular weight between the growing peptide and the
reaction by-products, while at the same time minimizing the
number of chain extension intermediates needing to be
resolved during UHPLC-MS reaction monitoring. The three
growing peptides are linked by an aromatic hub, which also
acts as an additional UV chromophore, by flexible, mono-
disperse octaethylene glycol (octagol, HO-Eg8-OH) spacers,
to limit steric interaction between adjacent peptides[13] (Fig-
ure 1). Compared to earlier MEPS strategies, this 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; and enhances the
separation due to the at least 3-fold mass difference between
Figure 1. Key features of PEPSTAR: Activated Fmoc-amino acid building blocks couple to chain termini of H-peptide-nanostar; piperidine
quenches excess amino acid and removes Fmoc; diafiltration washes all reaction by-products through a membrane; purified H-peptide-nanostar
is retained in the synthesizer, ready to repeat the cycle. Every process is analyzed in real time by UHPLC-MS.
Angew. Chem. Int. Ed. 2021, 60, 7786 –7795
ꢀ 2021 The Authors. Angewandte Chemie International Edition published by Wiley-VCH GmbH www.angewandte.org
7787

Angewandte
Research Articles
Chemie
nanostar and excess amino acids (AA). Furthermore, the
unimolecular, fully defined composition and UV chromo-
phore are commensurate with reaction monitoring by liquid
chromatography and mass spectrometry.
favourably with commercial supports (HO-Wang-PS
ꢀ 1.0 mmolgÀ1).[14]
H-Rink-nanostar 6 can be used directly in PEPSTAR
because the arms terminate with amino groups compatible
with standard peptide coupling chemistry. However, loading
the first amino acid onto a Wang-type anchor requires a slow
esterification, so this reaction was performed at high concen-
tration in a glass flask. We tested the esterification step with
large and sterically hindered Fmoc-Thr(tBu)-OH.[15] Fmoc-
Thr(tBu)-OH was condensed overnight with 4 using DcbCl
and HOBt in dichloromethane to give Fmoc-Thr(tBu)-O-
Wang-nanostar 7. Intermediate 7 was deprotected with DBU
and thiomalic acid to give loaded anchor H-Thr(tBu)-O-
Wang-nanostar 8, ready for PEPSTAR.
Nanostar anchor synthesis commenced from the known
tris-hydroxy terminated octagol-nanostar 1[12a,b] (Scheme 1a).
For compatibility with common peptide synthesis strategies
the termini must be modified. Either hydroxymethyl benzyl
groups, to mimic Wang resin, or the Rink amide linker were
appended to the chain termini of the nanostar. The hydroxyls
were first activated as their p-toluene sulfonyl esters 2, then
displaced by p-hydroxy benzaldehyde. The intermediate
tris(benzaldehyde) nanostar 3 was then reduced cleanly with
NaBH4 to provide the HO-Wang-nanostar 4. Alternatively,
the initial nanostar tosylate 2 was converted to the corre-
sponding triamine via Gabriel synthesis (phthalimide, then
hydrazinolysis, Scheme S2). The resulting tris(amino-octa-
gol)nanostar 5 was readied by amidation with Fmoc-Rink-
CO2H linker in the presence of N,N’-diisopropylcarbodiimide
(DIC) and hydroxybenzotriazole (HOBt), then deprotection
with 1,8-diazabicycloundec-7-ene (DBU) and thiomalic acid
(Scheme 1a); the latter quenches dibenzofulvene (DBF) to
form an adduct that is readily removed by aqueous wash-
ing.[5b] The resulting H-Rink-nanostar 6 (MW= 2120 Da) has
a loading capacity of 1.42 mmolgÀ1 which is higher than the
corresponding solid phase resin (H-Rink amide ChemMatrixꢀ
ꢀ 0.5 mmolgÀ1), or the 5 kDa PEG employed in MEPS
(0.2 mmolgÀ1). HO-Wang-nanostar 4 (MW= 1544 Da) has
a loading capacity of 1.94 mmolgÀ1 which also compares
Synthetic strategy. The widely used Fmoc-peptide chemis-
try was adopted for PEPSTAR.[16] In classical LPPS Fmoc-
peptide synthesis, four steps are typically required: Coupling,
isolation, deprotection and isolation again. Aiming for a one-
pot 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. This
has the advantages of reducing both solvent consumption and
total cycle time. A sacrificial species is required to react with
the active ester to prevent uncontrolled peptide chain
extension after Fmoc removal. Initially, in the 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 a UV-active by-product.
Unfortunately, the reaction of aniline with active esters was
Scheme 1. Synthesis of nanostars and chain extension cycle: a) Tris-hydroxy terminated octagol-nanostar 1 is converted to Wang-type, 4, and Rink,
6, nanostar; the 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 a three-step cycle of coupling, Fmoc removal, and diafiltration, then removed from the synthesizer for global
deprotection.
7788
www.angewandte.org ꢀ 2021 The Authors. Angewandte Chemie International Edition published by Wiley-VCH GmbH
Angew. Chem. Int. Ed. 2021, 60, 7786 –7795

Angewandte
Research Articles
Chemie
too slow to afford effective quenching. However, despite
OSN membrane and remained in the synthesizer. This
dispensing with the active ester quench after deprotection of
Fmoc-Asn(Trt)-Thr(tBu)-O-Wang-nanostar 9 with piperi-
dine, no double coupling contaminants could be detected in
the product H-Asn(Trt)-Thr(tBu)-O-Wang-nanostar 10 (Fig-
ure 2). We observed that remaining active ester was converted
instantaneously by excess piperidine into H-Asn(Trt)-Pip 11
(Scheme 1b).[17]
The reaction of excess AA with piperidine eliminated the
need for both an additional sacrificial species and a separate
diafiltration post-coupling. Thus, a synthesis 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: The 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; finally,
after UHPLC-MS confirmed complete deprotection, diafil-
tration commenced to remove the H-AA-Pip 11 derived from
excess AA, the DBF-Pip adduct 12, and other reaction by-
products, while the H-AA-nanostar was rejected by the
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 a peptide-nanostar proceeds to completion via two
intermediate species. Taking the formation of Fmoc-dipep-
tide-nanostar 9 as an example (Figure 2a), condensing [H-
Thr(tBu)-O -Wang]3-nanostar 8 with Fmoc-Asn(Trt)-OH, the
initial species formed has only one arm coupled, [Fmoc-
Asn(Trt)-Thr(tBu)-O-Wang]-[H-Thr(tBu)-O-Wang]2-nano-
star (1-arm, 13). The 1-arm intermediate is then consumed to
give a second species with two arms coupled (2-arm, 14), but
with one amino-terminated arm still free. Finally, the 2-arm
intermediate reacts on the remaining amino group to give the
desired Fmoc-dipeptide-nanostar 9. Each of these species is
chemically distinct and chromatographically resolvable. Be-
fore examining complex peptide-nanostars, the variation in
reaction rates for forming a range of dipeptides, and their
subsequent Fmoc removal, was examined in various solvents
(see SI section 4). Suspecting that a single solvent could not
Figure 2. Variation in reaction rates for a dimer-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 a general 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 1 wt%, blue
2 wt% and yellow 5 wt% starting concentration of H-Thr-nanostar 8. NB the drop in relative absorbance for 2 wt% after 125 mins in (e) was
most likely caused by sample preparation errors, high dilution was necessary in order to quench the reaction prior to the UHPLC-MS analysis.
Angew. Chem. Int. Ed. 2021, 60, 7786 –7795
ꢀ 2021 The Authors. Angewandte Chemie International Edition published by Wiley-VCH GmbH www.angewandte.org 7789

Angewandte
Research Articles
Chemie
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 a strong correlation between solvent
polarity and reaction rate; low 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, a THF-NMP 35:65 v/v mixture
was selected to provide a balance between reasonably fast
reaction kinetics, for both coupling and Fmoc removal, and
solubility, and was used for a study of the kinetics of peptide-
nanostar coupling and Fmoc removal.
Kinetic runs at three different concentrations of [H-
Thr(tBu)-O-Wang]3-nanostar 8 (1, 2 and 5 wt%) were
performed in a carousel reactor with THF-NMP 35:65 v/v
solvent (See SI section 5). Coupling reactions containing
1 wt% 8 and 5 equiv Fmoc-Asn(Trt)-OH (i.e. 1.7 equiv per
amino terminated arm) went to completion smoothly: the 1-
arm 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 4 hours (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, was reduced markedly (Figure 2e): at 2 and
5 wt% the coupling was finished within one hour and
2 minutes, 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, defining the
lower boundaries for nanostar concentration.
Membrane selection and synthesizer design. The OSN
membrane is at the heart of PEPSTAR technology. Three
chemically robust polymeric membranes developed in our
laboratory, capable of permeating larger solutes, were
screened for OSN: A polyethyleneimine (PEI) asymmetric
membrane, cross-linked with terephthalic chloride (TPC),
and coated with Jeffamineꢀ M-2005 (a polyether mono-
amine), denoted PEI_2005; two different batches of poly-
benzimidazole (PBI) asymmetric membrane, cross-linked
with a,a’-p-dibromoxylene (DBX), and modified with a poly-
mer brush Jeffamineꢀ 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] The key 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
a good solvent for peptide intermediates, and the rejections
were calculated based on Equation (S3) (Figure 3a, see also
Figure 3. Synthesizer design and separation performance: a) Rejection of peptide-nanostars 8 and 10, and 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); the early dip in the two-stage yield curve is a result of redistribution of a small proportion of 10 from stage 1 to stage 2.
Diavolume is the ratio of cumulative volume of wash solvent introduced to the synthesizer at any given time of diafiltration per volume of stage
1 (DV=Wash solvent volume / stage 1 volume), 1 DV=200 mL. c) Schematic of synthesizer layout, with picture of the synthesizer setup shown
in Figure S8. d) Purification of H-dipeptide-nanostar 10 in the synthesizer, from the largest by-products H-Asn(Trt)-Pip 11 and DBF-Pip 12; to the
2-arm by-product from diketopeparazine (DKP) can be detected, although this does not affect final peptide purity.
7790 www.angewandte.org ꢀ 2021 The Authors. Angewandte Chemie International Edition published by Wiley-VCH GmbH
Angew. Chem. Int. Ed. 2021, 60, 7786 –7795

Angewandte
Research Articles
Chemie
SI section 6). As expected from their MWs, H-Asn(Trt)-Pip
circulate at 90 LhÀ1, to minimize concentration polarization
11 always had a higher 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 A from B is expressed by
the separation factor, bA/B [Eq. (S4)] the higher it is the better
the separation. PBI_2005(1) gave the highest separation
factor b10/11 for purifying H-dipeptide-nanostar 10 from by-
product 11 (9.2) compared to both PBI_2005(2) (4.2) and
PEI_2005 (4.7); the separation factor for smaller H-AA-
nanostar 8 and by-product 11, b8/11, was again highest for
PBI_2005(1) (6.1). Although PBI_2005(2) gave a similar
peptide-nanostar rejection to PBI_2005(1), the separation
factor is much lower due to its simultaneously higher rejection
of by-product. Hence, PBI_2005(1) was selected for further
studies.
at the membrane surface.
The synthesizer was validated with a synthesis run to form
dipeptide-nanostar 10. The membrane cells of the synthesizer
were fitted with disks of PBI_2005(1), and stage 1 was charged
with H-Thr(tBu)-O-Wang-nanostar 8 in THF. On addition of
DIC, HOBt and Fmoc-Asn(Trt)-OH to the reactor, the
coupling proceeded smoothly to completion, as did the
subsequent Fmoc removal upon addition of 10%v/v piper-
idine. Once the chemistry was complete, the 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, and 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, as with its counterparts among SPPS supports, at i = 2 it
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% as determined on
chromatograms for example, Figure 3d), it introduces addi-
tional signals into the subsequent UHPLC chromatograms,
corresponding to (peptide-O-Wang)2-(HO-Wang)-nanostar,
that will confound the analysis of a new system. For this
reason, we changed to the H-Rink-nanostar anchor, used for
preparing C-terminal amides, because it does not produce
analogous DKP by-products. To validate the process of
repeated chain extension cycles during the development
phase of PEPSTAR, an enkephalin-like model peptide
sequence was designed (Figure 4a). The glycine residues of
the native Leu-enkephalin (H-Tyr-Gly-Gly-Phe-Leu-NH2)
sequence were replaced by serine to make (Ser2,Ser3)-Leu-
enkephalin pentapeptide (H-Tyr-Ser-Ser-Phe-Leu-NH2), be-
cause 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 1 of the synthesizer (total
ꢀ 200 mL) was charged with unloaded H-Rink-nanostar 6
( ꢀ 1.7 wt%) dissolved in THF-NMP 35:65. The same proto-
cols were then used to run the synthesizer, except that the
temperature was maintained at 358C for all reactions via
a heater 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 1 minute with DIC and
HOBt, was injected into stage 1 of the synthesizer via the feed
tank. Once UHPLC-MS confirmed completion of the cou-
pling, piperidine (20 mL, 10 v/v%) was added to the feed tank
to initiate Fmoc removal and quenching of the active ester;
Figure 3a shows that the rejection rises with MW, but only
approaches 100% rejection slowly. Thus, even though
PBI_2005(1) gave the highest separation factor, a significant
loss of yield is inevitable during diafiltration, particularly
during the early cycles of synthesis when the MW of the
peptide-nanostar is relatively low. For this reason, membrane
“cascade” systems were conceived that used membranes
linked in series to overcome the performance limits of a single
membrane by recycling partial flows back to previous stages.
Building on this concept, a practical two-stage separator,
requiring a single HPLC pump to pressurize the equipment,
and using pressure relief valves to control the pressure in each
stage, was implemented for the first time by Kim et al.[21]
Modelling, based on the measured rejection value for
PBI_2005(1), predicted that if a purity of 90% was required,
then at least 40% of dipeptide-nanostar 10 would be lost to
permeation using a single stage diafiltration system (Fig-
ure 3b). However, if a two-stage separation system was
adopted using the same membranes, yield loss improved to
about 10% while achieving the same purity. Our synthesizer
is based upon this two-stage design, which is a compromise
between achievable yield, complexity, and the time required
for diafiltration (which rises with the number of stages).
The key difference between the synthesizer described
here and earlier membrane separation apparatus[12] is that the
chemistry to assemble the heteropolymer, in this case
a peptide, occurs inside the same equipment as is used to
undertake purification of the crude product from reagent and
by-products. The principal elements of the synthesizer are
(Figure 3c): The feed-tank, into which fresh solvent is
passively added during OSN from a reservoir at the same
rate as the waste stream from OSN drains to a collection tank
(providing constant volume diafiltration); the 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 1 circulation loop containing two
membrane cells. The pressure is regulated by a pressure relief
valve that returns liquid from the stage 1 loop back to the feed
tank. The permeate from stage 1 enters the stage 2 circulation
loop where the pressure is again controlled by a pressure
relief valve. The stage 2 membrane concentrates the peptide-
nanostar permeating from stage 1 and a partial recycle flow
returns to stage 1, increasing yield. Pumps in both stages
Angew. Chem. Int. Ed. 2021, 60, 7786 –7795
ꢀ 2021 The Authors. Angewandte Chemie International Edition published by Wiley-VCH GmbH www.angewandte.org
7791

Angewandte
Research Articles
Chemie
Figure 4. Comparison of the synthesis of model peptides by PEPSTAR and SPPS: a) Synthesis of enkephalin-like model peptides. b) UHPLC
chromatograms for the chain extension of H-dipeptide-nanostar, exhibiting the 1-arm and 2-arm chain extended intermediates, and Fmoc removal.
c) Rise of H-peptide-nanostar rejection with peptide length; NB the error bars for stage 2 rejection (not shown) are large, due to low
concentrations and corresponding variation in peak integrals above noise. d & e) Comparison of the crude purities of pentapeptide 17 and
decapeptide 19 prepared by vendor SPPS or by PEPSTAR.
the piperidine was added slowly to further ensure that all the
AA was trapped before Fmoc removal commenced. As
before, diafiltration 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, Fig-
ure 4a).
All reactions and OSN were monitored in real-time with
UHPLC-MS. As before, the 1-arm and 2-arm intermediates,
as well as the 3-arm desired peptide-nanostar, are well
resolved on the chromatogram (see Figure 4b for i = 1 to 2),
permitting easy and sensitive reaction monitoring, and 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
1 hour, but 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. Monitoring of purification showed that
(e.g. Figure 3d) after 8 diavolumes (1 DV ꢀ 200 mL) had
permeated all large MW reaction by-products were removed,
including H-AA-Pip and DBF-Pip. However, purifications
were extended to 10 DV to reduce residual piperidine to
below 0.1 v/v% (as determined gas chromatography); this is
essential to prevent decomposition of fresh Fmoc-AA-OH
added at the start of the next cycle. As the peptide-nanostar
grew in length, its rejection also increased and reached
virtually 100% from i = 8 onwards (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, automatic 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, due 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-H2O 38:1:1 (Scheme 1b), after which
the crude pentapeptide H-Tyr-Ser-Ser-Phe-Leu-NH2 17 ex-
7792
www.angewandte.org ꢀ 2021 The Authors. Angewandte Chemie International Edition published by Wiley-VCH GmbH
Angew. Chem. Int. Ed. 2021, 60, 7786 –7795

Angewandte
Research Articles
Chemie
hibited 94% purity (Figure 4d). This level of purity, using
Octreotate. After validation of the PEPSTAR synthesis
only 1.7 equiv Fmoc-AA-OH per arm every cycle, is com-
parable with SPPS vendor using 3.0 equiv AA to obtain 97%
crude purity. Encouraged by this success, the H-pentapeptide-
nanostar 16 was further elongated to a decapeptide by
repeating the model (Ser2,Ser3)-Leu-enkephalin sequence
(Figure 4a). The same PEPSTAR protocols were applied,
with UHPLC-MS monitoring of reactions and OSN. At i = 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 deca-
peptide-nanostar 18, afforded crude decapeptide H-(Tyr-Ser-
Ser-Phe-Leu-)2-NH2 19 with 84% purity from 1.7 equiv,
considerably improving on the crude purity of 77% from
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, the 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, similar 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% to 35%. The impurities were
identified to be deletion sequences, presumably caused by on-
resin peptide aggregation or conformational effects during
the elongation from i = 5 to 10 which led to incomplete
coupling or Fmoc removal (Figure S12). However, the
authors emphasize that the solid phase synthesis with reduced
excess of AA was not optimized to achieve the highest
purities, but 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). Two additional runs were performed,
but with the initial H-Rink- nanostar 6 at 1 and 3 wt%, to
evaluate the effect of concentration on the performance of the
synthesizer. Under identical conditions to those above, the
1 wt% Rink anchor solution resulted in lower purities of
penta- and decapeptides 17 and 19 at values of 87% and 60%,
respectively; this is expected from slower kinetics as a result
of higher dilution. At 3 wt% 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 ꢀ 9 wt%, plus a further 3 wt% reagents,
gelation of the reaction solution occurred. As expected, any
LPPS process has a solubility ceiling that varies from one
target peptide to the next, but should be amenable to further
optimization of solvents, temperature and protecting groups.
For the reported model peptide, the optimum starting
concentration for H-Rink-nanostar 6 is around 2 wt%,
corresponding to 7 wt% of final H-decapeptide-nanostar
18.
system, we selected linear octreotate amide as a more typical
peptide target. Starting with an H-Rink-nanostar 6 concen-
tration of around 2 wt%, the octapeptide sequence was
synthesized using the same protocols as above in THF-NMP
35:65. Notably, the Fmoc-Cys(Acm)-OH building block was
selected, aiming for Cys to withstand the acidic global
deprotection with the Acm intact. All reactions and diafiltra-
tions went smoothly with no solubility issues. The 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, the degree of epimerization was also
investigated at the epimerization-prone Cys residue. The
global deprotection of H-Thr(tBu)-Cys(Acm)-Thr(tBu)-
Rink-nanostar (i = 3) showed < 0.1% of epimerization from
L- to D-Cys (Figure S14).
To further explore the capability of PEPSTAR, the
synthesis was recapitulated with H-Thr(tBu)-O-Wang-nano-
star 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 = 1 to 2 were 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 2 wt% and 2 v/v%. Diafil-
tration was then initiated, reducing the piperidine concen-
tration in stage 1, thus minimizing the time that piperidine
would have a high enough concentration to catalyze loss of
dipeptide to DKP. Even so, 10% (H-dipeptide-O-Wang)3-
nanostar was still converted to (H-dipeptide-O-Wang)2-(HO-
Wang)-nanostar via DKP. From i = 2 onwards, the standard
protocol was applied. All reactions and diafiltrations went
smoothly to obtain fully protected peptide-nanostar 22 in
71% yield. After global deprotection the purity of linear
octreotate was 70%, 88% if you include 18% cyclized
octreotate from loss of Acm, which is also comparable to
SPPS (Figure 5d, see also Figure S16). The fast reaction
kinetic facilitated in liquid phase was suspected to cause Acm
removal and promote cyclization. Hence, further optimiza-
tions 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 a putative
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 (kgkgÀ1).[23] The PMI evaluation shows that, for the
three methods under assessment, the solvent consumption
contributed to nearly all the total waste generated with only
a small fraction contributed by hubs (or resins) and other
reagents (see Table S6). The PMI is drastically lower for
PEPSTAR than for MEPS, and only slightly higher than for
SPPS. To compare production costs, we summed up the costs
of materials for PEPSTAR and MEPS to produce 1 mole of
linear octreotate amide based on current supplier prices
(Sigma–Aldrich). Using the same material prices, we then
Angew. Chem. Int. Ed. 2021, 60, 7786 –7795
ꢀ 2021 The Authors. Angewandte Chemie International Edition published by Wiley-VCH GmbH www.angewandte.org
7793

Angewandte
Research Articles
Chemie
Figure 5. Comparison of the synthesis of linear octreotate by PEPSTAR: a & b) Synthesis of linear octreotate amide and acid. c & d) Comparison
of the crude purities of SPPS vendor peptides with crude octreotate amide 21 and octeotate acid 23, respectively, prepared by PEPSTAR.
e) Comparison of Process Mass Intensity (PMI) and costs of materials for different methods of synthesis for octreotate amide 21.
estimated the cost for the vendor of SPPS to produce similar
quantity of product following the standard protocol (see SI
section 11). The cost 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, we anticipate
that the PMI and cost of PEPSTAR will fall below the
benchmark set by SPPS once downstream purification steps
are considered. The higher crude purity offered by liquid
phase synthesis in turn requires less demanding chromato-
graphic purification, which is often a major contributor to
waste generation.[24]
of PEPSTAR which, because of its entirely liquid phase
nature, is far less scale-limited than SPPS, is expected to
reduce the lengthy lead-times for process development, and
the PMI and production cost are anticipated to fall further at
larger scales. This technology is highly flexible in terms of
solvent and nanostar support choice. In future, full process
automation and further innovations on nanostar support and
solvent system will realize efficient synthesis of large peptides
or even proteins on PEPSTAR.
Acknowledgements
Conclusion
This work was supported financially by Eli Lilly through
a Lilly Research Award Programme grant and EPSRC
Programme Grant M01486/X. We thank Dr. Naveed Zaidi
for providing tris-hydroxy terminated octagol-nanostar. We
thank colleagues from Eli Lilly and Company for their
technical advice.
In this report, we have demonstrated liquid phase peptide
synthesis via one-pot nanostar sieving (PEPSTAR), a contin-
uous process to synthesize high purity peptides, without phase
or material transfers, suitable for automation. This advance
depended on the development of highly stable, polymeric
organic solvent nanofiltration (OSN) membranes that en-
abled reaction by-products to be efficiently “sieved” out from
the crude reaction mixture to fully purify the growing peptide
by diafiltration. Equally, the design of a three-armed star-
shaped monodisperse synthesis support, or nanostar anchor,
for growing peptides facilitated efficient membrane separa-
tion and real time monitoring, of both the synthesis and
purification phases of PEPSTAR. This one-pot synthesis
strategy lowered the PMI three-fold from its LPPS prede-
cessor, membrane-enhanced peptide synthesis (MEPS), and
is close to that of SPPS. The cost of materials is estimated to
be half of SPPSꢁs. This augurs well for manufacturing scale-up
Conflict of interest
The authors declare no conflict of interest.
Keywords: liquid phase peptide synthesis ·membranes ·
organic solvent nanofiltration (OSN) ·peptides
[1] a) A. P. Davenport, C. C. G. Scully, C. de Graaf, A. J. H. Brown,
J. J. Maguire, Nat. Rev. Drug Discovery 2020, 19, 389 –413; b) K.
7794 www.angewandte.org ꢀ 2021 The Authors. Angewandte Chemie International Edition published by Wiley-VCH GmbH
Angew. Chem. Int. Ed. 2021, 60, 7786 –7795

Angewandte
Research Articles
Chemie
Fosgerau, T. Hoffmann, Drug Discovery Today 2015, 20, 122 – [12] a) P. R. Gaffney, J. F. Kim, I. B. Valtcheva, G. D. Williams, M. S.
128.
Anson, A. M. Buswell, A. G. Livingston, Chem. Eur. J. 2015, 21,
9535 –9543; b) R. Dong, R. Liu, P. R. J. Gaffney, M. Schaeper-
toens, P. Marchetti, C. M. Williams, R. Chen, A. G. Livingston,
Nat. Chem. 2019, 11, 184 –184; c) J. F. Kim, P. R. J. Gaffney, I. B.
Valtcheva, G. Williams, A. M. Buswell, M. S. Anson, A. G.
Livingston, Org. Process Res. Dev. 2016, 20, 1439 –1452.
[13] M. Meldal, Methods in Enzymology, Vol. 289, Academic Press,
New York, 1997, pp. 83 –104.
[14] L. Ferrazzano, D. Corbisiero, G. Martelli, A. Tolomelli, A. Viola,
A. Ricci, W. Cabri, ACS Sustainable Chem. Eng. 2019, 7, 12867 –
12877.
[15] J. D. Young, A. S. Huang, N. Ariel, J. B. Bruins, D. Ng, R. L.
Stevens, Pept. Res. 1990, 3, 194 –200.
[16] W. C. Chan, P. D. White, Fmoc solid phase peptide synthesis:
a practical approach, Oxford University Press, Oxford, 2000.
[17] a) K. Ralhan, V. G. KrishnaKumar, S. Gupta, RSC Adv. 2015, 5,
104417 –104425; b) J. M. Collins, K. A. Porter, S. K. Singh, G. S.
Vanier, Org. Lett. 2014, 16, 940 –943.
[18] J. Lopez, S. Pletscher, A. Aemissegger, C. Bucher, F. Gallou,
Org. Process Res. Dev. 2018, 22, 494 –503.
[19] I. B. Valtcheva, P. Marchetti, A. G. Livingston, J. Membr. Sci.
2015, 493, 568 –579.
[20] A. M. C. Maan, A. H. Hofman, W. M. de Vos, M. Kamperman,
Adv. Funct. Mater. 2020, 30, 2000936.
[21] J. F. Kim, A. M. Freitas da Silva, I. B. Valtcheva, A. G. Living-
ston, Sep. Purif. Technol. 2013, 116, 277 –286.
[22] C. Goolcharran, R. T. Borchardt, J. Pharm. Sci. 1998, 87, 283 –
288.
[23] C. Jimenez-Gonzalez, C. S. Ponder, Q. B. Broxterman, J. B.
Manley, Org. Process Res. Dev. 2011, 15, 912 –917.
[24] a) L. Andersson, L. Blomberg, M. Flegal, L. Lepsa, B. L.
Nilsson, I. B. Valtcheva, Pept. Sci. 2000, 55, 227 –250; b) V.
Glaser, Genet. Eng. Biotechnol. News 2013, 33, 32 –34.
[2] a) M. Verlander, Int. J. Pept. Res. Ther. 2007, 13, 75 –82; b) A. J.
Mijalis, D. A. Thomas, M. D. Simon, A. Adamo, R. Beaumont,
K. F. Jensen, B. L. Pentelute, Nat. Chem. Biol. 2017, 13, 464 –
466.
[3] C. Hyde, T. Johnson, R. C. Sheppard, J. Chem. Soc. Chem.
Commun. 1992, 1573 –1575.
[4] a) D. Takahashi, T. Yamamoto, Tetrahedron Lett. 2012, 53, 1936 –
1939; b) D. Takahashi, T. Yano, T. Fukui, Org. Lett. 2012, 14,
4514 –4517; c) Y. Okada, H. Suzuki, T. Nakae, S. Fujita, H. Abe,
K. Nagano, T. Yamada, N. Ebata, S. Kim, K. Chiba, J. Org.
Chem. 2013, 78, 320 –327; d) J. Wu, G. An, S. Lin, J. Xie, W.
Zhou, H. Sun, Y. Pan, G. Li, Chem. Commun. 2014, 50, 1259 –
1261.
[5] a) Y. Okada, R. Takasawa, D. Kubo, N. Iwanaga, S. Fujita, K.
Suzuki, H. Suzuki, H. Kamiya, K. Chiba, Org. Process Res. Dev.
2019, 23, 2576 –2581; b) D. Takahashi, T. Inomata, T. Fukui,
Angew. Chem. Int. Ed. 2017, 56, 7803 –7807; Angew. Chem. 2017,
129, 7911 –7915; c) D. Takahashi (A.C. Inc.), WO/2012/029794,
2012; d) D. Takahashi (A. Co.), 2013; e) Y. Kono, H. Suzuki, S.
Moto (Jitsubo), WO2016140232A1, 2016; f) I. F. Eggen, F. T.
Bakelaar, A. Petersen, P. B. W. Ten Kortenaar, Org. Process Res.
Dev. 2005, 9, 98 –101.
[6] a) D. J. am Ende, Chemical Engineering in the Pharmaceutical
Industry : R and D to Manufacturing, Wiley, Chicester, 2010;
b) V. Castro, C. Noti, W. Chen, M. Cristau, A. Livignston, H.
Rodríguez, F. Albericio, Macromolecules 2017, 50, 1626 –1634.
[7] E. Bayer, M. Mutter, Nature 1972, 237, 512 –513.
[8] P. Marchetti, M. F. Jimenez Solomon, G. Szekely, A. G. Living-
ston, Chem. Rev. 2014, 114, 10735 –10806.
[9] a) S. So, L. G. Peeva, E. W. Tate, R. J. Leatherbarrow, A. G.
Livingston, Chem. Commun. 2010, 46, 2808 –2810; b) S. So, L. G.
Peeva, E. W. Tate, R. J. Leatherbarrow, A. G. Livingston, Org.
Process Res. Dev. 2010, 14, 1313 –1325.
[10] T. Zheng, M. Zhu, J. Yang, J. He, M. Waqas, L. Li, Macro-
molecules 2018, 51, 9333 –9343.
Manuscript received: October 28, 2020
[11] M. Paradís-Bas, J. Tulla-Puche, F. Albericio, Chem. Soc. Rev.
2016, 45, 631 –654.
Accepted manuscript online: January 14, 2021
Version of record online: February 24, 2021
Angew. Chem. Int. Ed. 2021, 60, 7786 –7795
ꢀ 2021 The Authors. Angewandte Chemie International Edition published by Wiley-VCH GmbH www.angewandte.org
7795