nature reviews chemistry https://doi.org/10.

1038/s41570-023-00493-y

Review article Check for updates

Synthesis and applications

of mirror-image proteins
Katriona Harrison , Angus S. Mackay
1,2
, Lucas Kambanis1,2, Joshua W. C. Maxwell1,2 & Richard J. Payne
1,2 1,2

Abstract Sections

The homochirality of biomolecules in nature, such as DNA, RNA, Introduction

peptides and proteins, has played a critical role in establishing and Accessing mirror-image
sustaining life on Earth. This chiral bias has also given synthetic proteins

chemists the opportunity to generate molecules with inverted chirality, Racemic protein X-ray
crystallography
unlocking valuable new properties and applications. Advances in the
field of chemical protein synthesis have underpinned the generation Mirror-image screening
technologies
of numerous ‘mirror-image’ proteins (those comprised entirely of
Mirror-image life
D-amino acids instead of canonical L-amino acids), which cannot be
accessed using recombinant expression technologies. This Review Conclusion and outlook

seeks to highlight recent work on synthetic mirror-image proteins,

with a focus on modern synthetic strategies that have been leveraged
to access these complex biomolecules as well as their applications
in protein crystallography, drug discovery and the creation of
mirror-image life.

School of Chemistry, The University of Sydney, Sydney, New South Wales, Australia. 2Australian Research Council
1

Centre of Excellence for Innovations in Peptide and Protein Science, The University of Sydney, Sydney, New South
Wales, Australia. e-mail: [email protected]

Nature Reviews Chemistry | Volume 7 | June 2023 | 383–404 383

Review article

Introduction early issues24. However, as D-amino acids are still more expensive than
The notion of chirality, first introduced by Louis Pasteur in 1848 (ref. 1), their L-counterparts, synthetic campaigns typically begin by testing
is used to describe the asymmetry of molecules, which have non- the synthetic strategy on the natural L-protein before applying the
superimposable ‘mirror-image’ forms2–4. Intriguingly, virtually all optimized protocol to the mirror-image molecule. The benefit of this
naturally occurring biomolecules on Earth display homochirality. approach is that the structure of the synthetic L-protein can be com-
That is, they are comprised of building blocks that all have the same pared with a recombinantly expressed version of the same protein,
chirality or ‘handedness’. The ribose sugars that make up strands of using techniques such as NMR spectroscopy and circular dichroism.
DNA and RNA are all D-configured, leading to the formation of their A direct comparison of the synthetic enantiomeric proteins can then
familiar right-handed α-helical structures. Similarly, the proteinogenic also be made to confirm that the mirror-image protein has the correct
amino acid monomers that make up polypeptides and proteins all con- 3D structure. It is important to note that following chemical synthesis,
tain L-configured chiral centres (apart from glycine which is achiral). D-proteins that possess disulfide bridges must also be subjected to
As well as posing fundamental questions about the basis for life, this folding, a step that can often require substantial optimization.
homochirality presents a range of unique opportunities5. Biomolecules The first synthesis of a mirror-image protein was reported in 1992
with inverted chirality compared with their natural counterparts (that by Kent and co-workers25, who succeeded in constructing both enanti-
is, the mirror-image) may be exploited for their distinctive properties, omers of the 99-residue HIV-1 protease enzyme entirely en bloc using
particularly in chemical reactions and biological systems. For example, tert-butyloxycarbonyl (Boc)-solid-phase peptide synthesis (SPPS)
D-proteins are known to be resistant to proteolytic degradation, as (Fig. 1a). This pioneering work not only kickstarted the field of mir-
native proteases (themselves made up of L-amino acids) are only capa- ror-image protein synthesis, but was also the first to demonstrate
ble of recognizing and cleaving proteins comprised of L-amino acids. chiral specificity between an enzyme and its chiral peptide substrate.
These unique properties mean that mirror-image biomolecules may Specifically, the D-protease was only effective at cleaving D-peptides
be used for drug discovery and therapeutic development, improved and, notably, was not capable of cleaving the enantiomeric L-peptides.
protein crystallography manifolds and, ambitiously, the construction As expected, an achiral inhibitor had no preference for one enantio-
of mirror-image life. meric form of the protein over the other. In addition to presenting the
earliest example of the chemical synthesis of a mirror-image protein,
Accessing mirror-image proteins this seminal study foreshadowed the widespread application of
For decades, recombinant techniques have relied on the use of D-proteins in protein science and drug discovery.
genetically engineered microorganisms, such as the bacterium More recently, Kay and co-workers26 successfully synthesized
Escherichia coli, to produce proteins on a large scale in an affordable L-configured and D-configured versions of the 312-residue E. coli DapA
manner6,7. These revolutionary technologies have enabled access to protein (4-hydroxy-tetrahydrodipicolinate synthase) to investigate
both native and designer proteins with relative ease and allowed for the whether the GroEL–ES chaperone system could be used to fold mirror-
effect of individual amino acid residues on protein structure and func- image proteins (Fig. 1b). E. coli DapA protein was selected as a suitable
tion to be investigated through site-specific mutagenesis8. However, as target as it is typically found in cells that are highly enriched in GroEL–ES
recombinant methods are reliant on the ribosomal machinery of the complexes, which help to prevent aggregation, degradation and, by
host organism for protein synthesis, only a relatively narrow chemical extension, the loss of activity of this protein. To generate a target of
space can be explored, with the primary sequence generally limited to this size, fluorenylmethyloxycarbonyl (Fmoc)-SPPS was first used
combinations of the proteinogenic L-amino acids9. In nature, the pro- to produce seven peptide fragments that were then ligated together
teome is imbued with further complexity through the introduction of using a convergent native chemical ligation (NCL) strategy leveraging
protein post-translational modifications (PTMs), including chemoen- acyl hydrazide to thioester conversion chemistry. Of note, alanine
zymatic conversion of specific residues from the native L-configuration (Ala)77 (located at the N-terminus of the third peptide fragment) was
to the mirror-image D-form10,11. More recent developments have also mutated to a cysteine (Cys) residue to remove the need for an additional
allowed for the incorporation of non-native functionality in recombi- desulfurization step, thereby helping to maximize product recovery
nant systems using innovative genetic reprogramming techniques such at the typically low-yielding histidine–Ala junction. Importantly, the
as amber codon suppression for the introduction of D-amino acids12. authors first analysed the crystal structure of native L-DapA, which
However, accessing complete mirror-image proteins, in which every revealed that Ala77 was situated away from the protein active site.
residue is D-configured, remains challenging, and chemical protein This suggested that the mutation would cause minimal disruption to
synthesis, rather than recombinant expression, remains the only means the protein structure and function. Following assembly via an iterative
to access these mirror-image biomolecules. Crucially, this approach, ligation approach, both synthetic L-DapA and D-DapA A77C mutants
whereby each amide bond in a protein of interest is constructed using were denatured and incubated with GroEL–ES for protein folding. After
synthetic chemistry, is compatible with D-amino acid building blocks13. folding, the enzymatic activities of synthetic L-DapA and D-DapA were
There have been numerous important developments in the field of 60% and 40%, respectively, relative to the activity of GroEL–ES-folded
chemical protein synthesis14–22 that have underpinned the generation recombinant wild-type DapA. This pioneering work revealed that the
of mirror-image proteins (outlined in Box 1). native GroEL–ES system was, surprisingly, able to fold a D-protein.
The ambidextrous folding ability of this chaperone removes a substan-
Case studies of mirror-image protein synthesis tial barrier to the construction of large mirror-image protein systems
Research on mirror-image proteins was initially hampered by a lack of and should stimulate further research into the chiral specificity (or lack
chemical tools for accessing larger polypeptides and the prohibitive thereof) of other cellular machinery.
cost of D-amino acid building blocks23. The advent of ligation technol- Kay and co-workers27 have also reported the synthesis of the
ogy and developments in the industrial syntheses of D-amino acids from 52-kDa soluble domain of D-tumour necrosis factor alpha (D-TNFα)
cheap and readily available precursors have helped to overcome these (Fig. 1c). TNFα is an inflammatory cytokine implicated in many chronic

Nature Reviews Chemistry | Volume 7 | June 2023 | 383–404 384

Review article

Box 1

Tools for chemical protein synthesis

Solid-phase peptide synthesis
Solid-phase peptide synthesis (SPPS) relies on an insoluble polymer
a SPPS of D-polypeptides R1
O Resin
support (resin) that provides an anchor to which successive amino R1 H2N
cleavage O R O
H
acids are coupled to generate a polypeptide chain159. Coupling O O and global H2N
N
N
OH
PGHN Deprotection deprotection
proceeds via a condensation reaction between the α-carboxylic O R1
H
O
n
R n+2
acid of an N-terminally protected amino acid and the N-terminal O
O
R1 PGHN
α-amine of the growing resin-bound peptide chain to form a new O Coupling
O
2 H2N Peptide OH
PGHN R
amide bond (see the figure, part a). Although SPPS is a versatile reagents
Coupling reagents
O O R1
platform, the method is limited by the size of the peptide fragments
PGHN O
X
that can be accessed (typically ~50 residues). To chemically N
H
R2 O
synthesize larger proteins, additional synthetic, semi-synthetic
and chemoenzymatic methods have been developed160–164. b NCL O H2N Peptide 1 O
H2N Peptide 1 SR Transthioesterification O
S
Native chemical ligation HS O Peptide 2 OH
The advent of native chemical ligation (NCL), a ground-breaking Peptide 2
H2N
OH O
methodological advance, overcame the length restrictions of H2N S to N
O acyl shift
SPPS165. This convergent ligation procedure joins two unprotected O
O
Desulfurization
HS O
peptide fragments in a traceless manner, generating a native peptide Peptide 2 OH O
H2N Peptide 1 N
bond. An initial intermolecular transthioesterification between the H
O H2N Peptide 1 N
Peptide 2 OH
H
N-terminal cysteinyl thiol of one peptide fragment and the C-terminal O

thioester of a second peptide fragment joins the fragments together c Peptide hydrazide ligation
via a thioester linkage (see the figure, part b). An intramolecular S→N O i. NCL
ii. Acyl
acyl shift of this intermediate is entropically favoured by nucleophilic H2N Peptide 1 SR hydrazide O
HS O

attack of the α-amine moiety onto the thioester, leading to the to thioester
H2N Peptide 1 N
Peptide 2 SR
HS O
formation of a native peptide (amide) bond161. H
O
Peptide 2 NHNH 2 HS O
H2N
Peptide 3
Desulfurization O H2N
OH

NCL
As cysteine (Cys) has a relatively low natural abundance in proteins HS
O
O
HS O
(1.4%), the need for one of the fragments to possess an N-terminal O
Peptide 3 OH
Peptide 2
Cys residue greatly restricts the number of ligation junctions that H2N Peptide 1 N N
H
H O
are accessible using NCL. To overcome this, a desulfurization O

strategy was developed whereby the NCL product is treated with d C-terminal to N-terminal iterative ligation
O
the reductant tris(2-carboxyethyl)phosphine, the water-soluble i. NCL
radical initiator 2,2ʹ-azobis[2-(2-imidazolin-2-yl)propane] (VA-044) R′ Peptide 2 SR ii. Protecting
HS O
group HS O
and a hydrogen atom source (such as tBuSH)166. Several additional O removal
Peptide 3 OH
HS O Peptide 2
desulfurization methodologies have since been reported and applied H2N
N
H
Peptide 3 OH O
to the synthesis of protein targets167,168. H2N O
O
O
R′ = H2N Peptide 1 SR
Peptide hydrazide ligation S AcmS NCL
HS O
The use of peptide acyl hydrazides as thioester surrogates has N
or HS O
H2N O
H Peptide 3
become a popular strategy for the assembly of proteins in the Peptide 2 N
OH
H2N Peptide 1 N H
N-terminal to C-terminal direction169 (see the figure, part c). Activation H
O
O

using NaNO2 or acetylacetone generates the corresponding acyl

azide or acyl pyrazole, respectively, with subsequent thiolysis
affording a C-terminal thioester that can be ligated to another peptide to convergent syntheses in which iterative ligations are required to
fragment via NCL170,171. Importantly, this chemoselective approach yield the protein of interest161,172. In such cases, orthogonal-protecting
is compatible with peptide fragments constructed using SPPS as groups (such as acetamidomethyl (Acm)) are required to mask
well as recombinantly accessed intein fusion proteins160. Therefore, N-terminal Cys residues during the hydrazide ligation, to avoid
the introduction of peptide hydrazides as masked thioesters has undesired intramolecular cyclization or polymerization164. Protecting
expanded the range of targets accessible using chemical protein group removal post-ligation liberates the reactive N-terminal Cys
synthesis and semi-synthesis. As the acyl hydrazide functional group residue, enabling subsequent ligations to be performed in the
is unreactive under NCL conditions before activation, it is well suited C-terminal to N-terminal direction (see the figure, part d).

Nature Reviews Chemistry | Volume 7 | June 2023 | 383–404 385

Review article

a Syntheses of HIV-1 protease enantiomers

Mirror plane
O O
BocHN NHBoc
O O O O

N N
H H

Boc L -SPPS Boc D-SPPS

O O
Folding Mirror-image Folding
HIV-1 protease– H2N D-HIV PR (1–99) OH
H2N L-HIV PR (1–99) OH HIV-1 protease–
substrate complex
substrate complex

b Synthetic strategy for the assembly of D-DapA HS O HS O

O D-DapA (78–119) NHNH2 D-DapA (121–160)

H2N NHNH2
i) Acyl hydrazide H2N
H 2N D-DapA (1–39) to thioester O 77 O 120 O
NHNH2 O O
ii) NCL
HS H2N D-DapA (1–76) NHNH2 H2N D-DapA (1–160) NHNH2
O H2N D-DapA (1–119) NHNH2
i) Acyl hydrazide i) Acyl hydrazide
D-DapA (41–76) to thioester to thioester
NHNH2
H2N ii) NCL ii) NCL
40 O
HS O
AcmS i) Acyl hydrazide
O
i) Acyl hydrazide D-DapA (239–312) NH2 to thioester
to thioester H2N ii) NCL
D-DapA (162–210) NHNH2
H2N ii) NCL AcmS O 238 O HS O
iii) Desulfurization
161 O
D-DapA (162–237) NHNH2 D-DapA (162–312) NH2
H2N i) Acyl hydrazide H2N
HS O 161 O to thioester 161 O O
ii) NCL
D-DapA (212–237) NHNH2 iii) Acm deprotection H2N D-DapA (1–312) NH2
H2N
211 O Ambidextrous
GroEL–GroES
complex folding

D -DapA tetramer

c Synthesis of biotinylated D-TNFα

O
i) Acyl hydrazide
AcmS Acm O to thioester Biotin-PEG2 D-TNFα (78–144) NHNH2
N 77
ii) NCL H
D-TNFα (146–184) NHNH 2 iii) Desulfurization O O
H2N HS O
iv) Acm deprotection
145 O D-TNFα (146–233) NH2 Biotin-PEG2 D-TNFα (78–233) NH 2
H2N i) Acyl hydrazide N 77
HS O H
145 O to thioester O
ii) NCL
D-TNFα (186–233) NH2
H2N iii) Oxidative folding

185 O

d Convergent synthesis of biotinylated KRas(G12V)

S O
t
BuSS Alloc O CO2Et
i) Thioacid activation D-KRas (81–117) S
N
D-KRas (119–138) and ligation H
SH 80 O HS O
FmocHN ii) Fmoc deprotection t
BuSS O
iii) Alloc deprotection PEG2-biotin
118 O PEG2-biotin D-KRas (81–166) N
D-KRas (119–166) N i) NCL H2N H
H2N H 80 O
Alloc Alloc O ii) Thiazolidine
118 O removal
PEG2-biotin
D-KRas (140–166) N
H2N H
H
O
139 i) NCL
ii) Hmb removal
MeO OH Hmb Hmb O CO2H
Hmb Hmb O
H2N D-KRas (1–50) S NCL CO2Et
O H2N D-KRas (1–79) S
Hmb = N
HS O
O CO2Et O
Disrupts peptide secondary D-KRas (52–79) S
H2N PEG2-biotin
structure for improved stability H2N D-KRas (1–166) N
51 O H

Nature Reviews Chemistry | Volume 7 | June 2023 | 383–404 386

Review article

Fig. 1 | Synthetic strategies for mirror-image proteins. a, Syntheses of HIV-1 DapA corresponds to the structure of L-DapA (PDB 1DHP) at 2.5 Å resolution158.
protease (PR) enantiomers using either D-tert-butyloxycarbonyl (Boc) or c, Synthesis of biotinylated D-tumour necrosis factor α (TNFα) (77–233) via
L-Boc-phenylalanine-OCH2-polyacrylamide resin. Following iterative cycles iterative acyl hydrazide to thioester conversions and native chemical ligation
of Boc-solid-phase peptide synthesis (SPPS), the full-length proteins were (NCL). d, Convergent synthesis of biotinylated KRas(G12V) utilizing a tert-butyl
cleaved off resin and folded (PDB 1HPX). b, Synthetic strategy for the assembly isocyanide-mediated thioacid activation fragment condensation approach. Acm,
of D-DapA (4-hydroxy-tetrahydrodipicolinate synthase). The synthetic D-protein acetamidomethyl; Alloc, allyloxycarbonyl; Fmoc, fluorenylmethoxycarbonyl;
was folded using a GroEL–ES chaperone system. Note: Ribbon structure of Hmb, 2-hydroxy-4-methoxybenzyl; PEG, polyethylene glycol.

autoimmune diseases such as rheumatoid arthritis28. The discovery of protein X-ray crystallography, readers are directed to a comprehensive
high-affinity ligands to this protein is therefore an attractive avenue review by Yeates and Kent48.
for the development of novel therapeutics. The synthetic strategy In 1989, Mackay49 first discussed using a mixture of a pair of protein
involved the generation of three D-peptide fragments via Fmoc-SPPS. enantiomers to crystallize into space group P1 <bar>, the most com-
D-TNFα (145–184) bearing an N-terminal acetamidomethyl (Acm)- monly observed of 165 possible achiral space groups. Formation of
protected Cys residue and a C-terminal acyl hydrazide was converted to a centrosymmetric crystal from a racemic protein mixture, wherein
the thioester and ligated to D-TNFα (185–233) using NCL. Following des- both enantiomers are inverted on one another in the unit cell, restricts
ulfurization and Acm deprotection, the D-TNFα (145–233) segment was the phases to either 0o or 180o (as opposed to any value from 0o to
ligated to D-TNFα (77–144) thioester (generated from the corresponding 360o). Given that native L-proteins are homochiral, they can only access
acyl hydrazide) to afford the target mirror-image TNFα protein. space groups that are also chiral and hence cannot form centrosym-
The total synthesis of L- and D-forms of a 166-residue oncogenic metric crystals. By mixing a native L-protein with its corresponding
protein, mutant KRas(G12V), has also been reported with a view for D-enantiomer (produced synthetically) and performing racemic X-ray
application in mirror-image ligand screening29 (Fig. 1d). Assembly crystallography, the complications of phase determination are drasti-
of the five peptide fragments was achieved using a combination of cally reduced owing to limitations in the number of centrosymmetric
NCL-desulfurization and fragment condensation via an isocyanide- space groups that are available. Following this report, Zawadzke and
mediated activation strategy30. Four suitable ligation junctions were Berg50 synthesized both enantiomers of the 45-residue peptide rubre-
identified, with each fragment constructed via Fmoc-SPPS. Activation doxin for racemic X-ray crystallography. The authors demonstrated
of the C-terminal thioacid at residue 138 was performed using tert-butyl that the crystal from the racemic mixture was centrosymmetric. High-
isocyanide to induce ligation with the N-terminal isoleucine (Ile) resi- quality electron density maps were obtained and the structures of
due of KRas(139–166). A series of convergent NCL reactions completed both enantiomers were solved to 2 Å resolution by molecular replace-
the synthesis. Of note, attaching 2-hydroxy-4-methoxybenzyl (Hmb) ment. Higher numeric values of resolution (>4 Å) are considered poor
groups to the amide backbone at Gly10 and Gly15 markedly reduced resolution, whereas lower values (≤2 Å) are considered high resolution.
handling and solubility issues for the KRas(1–50) fragment. After
folding, circular dichroism and size-exclusion chromatography were Racemic protein X-ray crystallography uses
used to show that L-KRas(G12V) displayed structural features similar Racemic protein X-ray crystallography has rapidly risen to become a key
to recombinantly expressed KRas(G12V). A diethyleneglycol-biotin tool for elucidating structures of proteins that are otherwise resistant to
handle was incorporated for both D-TNFα and D-KRas(G12V). This crystallization. A particularly interesting example was reported by Kent
will enable the proteins to be immobilized on streptavidin beads for and co-workers who synthesized both enantiomers of the 94-residue
use as bait in mirror-image screening campaigns. A critical avenue Mycobacterium tuberculosis (Mtb) protein Rv1738 using NCL51 (Fig. 2b).
for future research will be the use of these, as well as other recently The gene that encodes Rv1738 is the most upregulated gene when Mtb
prepared synthetic mirror-image domains of oncogenic proteins (such is in its dormant state, meaning that Rv1738 is a protein of considerable
as SH2 (refs. 31,32) and Ig2 Axl33), in mirror-image display campaigns interest, particularly as the dormant bacteria do not respond well to
for elucidating high-affinity ligands. Exciting new advances in the flow- current frontline treatments52,53. The authors made extensive attempts
based solid-phase synthesis of D-proteins have the potential to rapidly to crystallize the natural L-Rv1738 protein, without success. However,
provide many more mirror-image targets for therapeutic discovery via using a racemic mixture of synthetic D-Rv1738 and L-Rv1738 proteins,
display technologies34. crystals were obtained within days and the structures were solved for
both enantiomers, to a resolution of 1.5 Å. Surprisingly, analysis of the
Racemic protein X-ray crystallography crystal structure revealed high structural homology between Rv1738
Mirror-image proteins can be used to help determine the crystal struc- and several bacterial hibernation proteins, despite limited primary
tures of native L-proteins using an analytical technique called racemic sequence homology. This intriguing discovery exemplifies how mirror-
protein X-ray crystallography (Fig. 2a). Obtaining X-ray crystal struc- image proteins in racemic crystallography can be used to advance our
tures of proteins can be difficult owing to the challenging nature of, understanding of the structure and function of native proteins where
first, generating crystals of diffraction quality and, second, undertaking other biochemical techniques have failed.
phase determination to solve the final crystal structure35. Combining Racemic protein X-ray crystallography has also been used to
a native L-protein with an equal quantity of its enantiomeric D-protein uncover key structural information about other biologically important
can help to overcome these barriers and enable the facile formation proteins, for example, to solve the first crystal structure of Calcisep-
of well-ordered crystals. Since its inception, this technique has been tine (CaS), a potent neurotoxin found in Dendroaspis polylepis (black
critical for solving the crystal structures of many recalcitrant L-proteins mamba) venom54. Synthetic L-CaS and D-CaS were assembled using
and has become a widely used technique in the field of protein crystal- an acyl hydrazide-based NCL strategy, followed by oxidative folding
lography36–47. For a more detailed background on the theory of racemic to form the four native disulfide bonds (Fig. 2c). The crystal structure

Nature Reviews Chemistry | Volume 7 | June 2023 | 383–404 387

Review article

a Racemic/quasi-racemic mixture protocol b Synthesis and racemic X-ray crystallography of L-Rv1738 and D-Rv1738 O
Modified L-protein D-protein HS O
i) NCL H 2N L-Rv1738 (1–29) SR
HS O
L-Rv1738 (67–94) OH ii) Thz HS O
H2N deprotection
66 H2N L-Rv1738 (1–94) OH
O L-Rv1738 (31–94) OH i) NCL
Point H2N
O 30
mutation S O
ii) Desulfurization
L-Rv1738 (31–65) SR
N
H 30
O Racemic X-ray
crystallography
O
HS O L-Rv1738
i) NCL H2N D-Rv1738 (1–29) SR
HS O
D-Rv1738 (67–94) OH ii) Thz HS O
1 H2N deprotection
Quasi-racemic Racemic 66 H2N D-Rv1738 (1–94) OH
O D-Rv1738 (31–94) OH
protein mixture protein mixture H2N i) NCL
S O 30 ii) Desulfurization
O
Crystallization N
D-Rv1738 (31–65) SR
H 30
O
Protein
sample
c Synthesis and racemic X-ray crystallography of L-CaS and D-CaS
O
i) Acyl hydrazide
Crystallization H2N L-CaS (1–38) NHNH2 to thioester O
mix ii) NCL
HS O H2N L-CaS (1–60) NH2
2
L-CaS (40–60) NH2
X-ray diffraction H2N Folding
L-CaS
39
O

Racemic X-ray
crystallography
Protein
crystal O
i) Acyl hydrazide Folding
H2N D-CaS (1–38) NHNH2 to thioester O FS2
ii) NCL
HS O H2N D-CaS (1–60) NH2

D-CaS (40–60) NH2

Analysis H2N
39
O
L-Protein

d Synthesis and racemic X-ray crystallography of rC5a-desArg

O
i) Acyl hydrazide
H2N L-rC5a-desArg (1–23) NHNH2 to thioester O L-rC5a-desArg
ii) NCL
HS O H2N L-rC5a-desArg (1–76) NH2

L-rC5a-desArg (25–76) NH2 Folding

H2N
24 Major
O structural
Racemic X-ray difference
O crystallography
i) Acyl hydrazide Folding
H2N D-rC5a-desArg (1–23) NHNH2 to thioester O
ii) NCL
HS O H2N D-rC5a-desArg (1–76) NH2

D-rC5a-desArg (25–76) NH2

H2N Native L-hC5a
24
O

e Synthesis and enhanced quasi-racemic X-ray crystallography

i) Acyl hydrazide to thioester
HS O ii) NCL O
O
iii) Desulfurization
+ D-Ub (47–76) NH2 H2N D-Ub (1–76) NH2
H2N D-Ub (1–45) NHNH2
H2N
46 Monomeric D-ubiquitin
O
Monomeric D-ubiquitin Monomeric D-ubiquitin

180º
Rotation

Trimeric L-ubiquitin Tetrameric L-ubiquitin

Nature Reviews Chemistry | Volume 7 | June 2023 | 383–404 388

Review article

Fig. 2 | The racemic and quasi-racemic protein X-ray crystallography (Fmoc)-SPPS were ligated by NCL using acyl hydrazide to thioester conversion
workflow and application for accelerating crystallization of synthetic chemistry. Representation of crystal structure of L-CaS (PDB 2ERA) coloured in
proteins. a, Racemic/quasi-racemic mixture protocol. (1) Crystallization: yellow compared with related protein family member L-FS2 in purple (PDB 1TFS).
1:1 protein mixtures of native and mirror-image proteins with either exact d, Synthesis and racemic X-ray crystallography of rC5a-desArg from two peptide
primary sequence and post-translational modifications (racemic) or slight fragments, synthesized by Fmoc-SPPS, and ligated by NCL via acyl hydrazides as
structural variation (quasi-racemic) are exposed to crystallization mix, which thioester precursors. Representation of racemic crystal structure of rC5a-desArg
enhances the likelihood of crystallization. (2) X-ray diffraction: Crystals of a (adapted from PDB 4P3A to mimic structure) (yellow), compared with native
1:1 protein mixture are subjected to X-ray crystallography, and crystal structures hC5a (PDB 1KJS, NMR structure) (blue). e, Synthesis and enhanced quasi-racemic
of the enantiomeric protein pair are solved through analysis of the diffraction crystallography of ubiquitin. Two peptide fragments, synthesized by Fmoc-SPPS,
pattern. b, Synthesis and racemic X-ray crystallography of L-Rv1738 and were ligated by NCL (following acyl hydrazide to thioester conversion) with a
D-Rv1738 via tert-butyloxycarbonyl (Boc)-solid-phase peptide synthesis (SPPS) subsequent desulfurization step. Quasi-racemic crystal structures of M1-linked
of three peptide fragments, ligated using a native chemical ligation (NCL)- tri-ubiquitin (adapted from PDB 5GO7) or tetra-ubiquitin (adapted from PDB
desulfurization strategy. Racemic X-ray crystallography led to the determination 5GO8) (purple) with monomeric D-ubiquitin (blue). R indicates an alkyl thioester;
of a crystal structure of L-Rv1738 (PDB 4WPY) (red). c, Synthesis and racemic rC5a-desArg, rat complement fragment 5a with C-terminal arginine deletion;
X-ray crystallography of L-CaS and D-CaS (snake toxin Calciseptine). Two Thz, thiazolidine.
L-peptide and D-peptide fragments synthesized by fluorenylmethoxycarbonyl

was solved to 1.69 Å resolution, revealing that the toxin possessed characterized by an extensive α-helical network and a heterochiral
an interesting three-finger structure. Importantly, this information interaction between L-melittin and D-melittin homochiral dimers. The
allowed the structure of CaS to be compared with other proteins in the choice of melittin was strategic, as it was a rare case in which the crystal
same family and provides a key resource for rational structure-based structure of an L-configured membrane protein had previously been
drug design in the future. obtained using single X-ray crystallography60. A detailed comparison
A similar synthetic approach was applied to probe the structure showed that the racemic and L-structures were in close agreement,
of modified rat complement fragment 5a (rC5a), which has been impli- suggesting that racemic crystallography offers a useful approach
cated in a range of autoimmune diseases55. Interestingly, circulating for determining the native quaternary structure of peptides and pro-
C5a is rapidly inactivated by removal of the C-terminal arginine (Arg) teins and holds promise for improving the ease of deciphering native
to form C5a-desArg56. Chemical synthesis of both enantiomers of rC5a polypeptide quaternary structures.
with C-terminal Arg deletions (rC5a-desArg) was followed by racemic The utility and importance of racemic X-ray crystallography are
X-ray crystallography, with the structure solved to 1.8 Å resolution evidenced by its widespread adoption for the determination of peptide
(Fig. 2d). A substantial structural change at the C-terminus of rC5a- and protein structures. As the expansion of the field has coincided with
desArg was observed compared with the previously reported human improvements in strategies for accessing proteins by chemical synthe-
C5a (hC5a) structure, providing the first structural evidence for the sis, the technique is now considered routine. In addition to their role
observed decrease in affinity57. Using these insights, it was proposed in racemic crystallography, synthetic mirror-image proteins can also
that removal of the positively charged Arg induces a pronounced con- be applied in a related technique called quasi-racemic protein X-ray
formational change at the C-terminus, leading to the different receptor crystallography, which has expanded the suite of proteins that can be
binding affinities of C5a and C5a-desArg. crystallized and crystal structures resolved.
The first crystal structure of a β-sheet antimicrobial θ-defensin
(namely, BTD-2) was also obtained using racemic X-ray crystallography, Quasi-racemic protein X-ray crystallography
thus enabling investigation into the relationship between oligomeric Alongside racemic protein mixtures, pairs of proteins that are not
structure and bacterial toxicity58. BTD-2 was accessed via SPPS, with exact mirror-images but are of a similar shape (called enantiomorphs)
a racemic mixture yielding crystals that enabled the structure to be can also promote crystal formation48,61. This concept is referred to
solved at 1.45 Å. Excitingly, this work uncovered a novel oligomeric as quasi-racemic crystallography. In the context of proteins, a single
form of β-sheet antimicrobial peptide, termed an anti-parallel trimer, D-protein may be used to help determine the crystal structure of a series
which provided key evidence for the controversial hypothesis that of L-analogues, which display small structural differences compared
this is the membrane-active form of BTD-2 and other similar defensin with the native L-protein62. The atomic structure of the D-protein is also
proteins. The authors noted that this family of β-sheet peptides, which determined during this process and may be inverted to generate the
form amyloid-like fibrils, are related to amyloid-forming peptides, with crystal structure of the parent L-protein. This means that the structures
their crystal structure providing the first evidence that the two families of the parent L-protein and the L-analogues may be directly compared
of peptides bear structural similarities. under identical crystallization conditions, without needing to synthe-
Racemic crystallography has also been used to characterize native size the native L-protein. As such, this technique (similar to racemic
polypeptide quaternary structures. Overcoming widespread reserva- crystallography) may be used to overcome issues with crystallizing
tions about this approach, crystals of a racemate of the membrane- difficult proteins, while also serving to reduce the number of protein
active peptide melittin enabled its structure to be solved at 1.27 Å, targets that need to be synthesized63.
in which a tetrameric assembly was observed59. This was a notable Quasi-racemic protein crystallography was first reported in 2008,
advance, as it was unclear whether racemic crystallography could following the total synthesis of an 81-residue snow flea antifreeze pro-
determine native quaternary structures (which can be disrupted by tein (sfAFP), which had no known homologues and could not be crystal-
heterochiral associations and are considered resistant to crystalli- lized in its native form64,65. Although a racemic mixture of L-sfAFP and
zation). Intriguingly, the structure revealed a tetrameric assembly D-sfAFP gave diffraction-quality crystals, improved crystallization was

Nature Reviews Chemistry | Volume 7 | June 2023 | 383–404 389

Review article

observed when using a quasi-racemic mixture comprising of L-Se-sfAFP solved at high resolution (0.90–1.22 Å). All three analogues had the
and D-sfAFP. L-Se-sfAFP was generated by replacing asparagine111 same final folded structure as wild-type ShK toxin and retained Kv1.3
in L-sfAFP with a selenocysteine residue that was alkylated with bro- ion channel blocking activity with IC50 values ranging from 440 pM to
moacetamide. Site-specific introduction of an anomalous scattering 860 pM. D-allo-ShK and L-allo-ShK were also synthesized; however,
atom (Se) into one enantiomer facilitates phase determination in quasi- neither of them folded, contradicting previous reports72. Ultimately,
racemic X-ray crystallography. This approach led to crystal formation, it was concluded that the initial D-allo-ShK reported was actually D-ShK
which enabled the structure to be solved to 0.98 Å resolution. and that the observed activity was likely caused by contamination by
Quasi-racemic crystallography was also used to investigate the the native L-ShK toxin74.
influence of glycosylation on the structure of the Ser-CCL1 chemokine66. As for racemic protein X-ray crystallography, quasi-racemic pro-
Despite glycosylation being a common PTM, it is challenging to obtain tein X-ray crystallography is a highly valuable technique for facilitating
crystal structures of glycosylated proteins owing to the heterogeneous protein structure determination and the elucidation of structure–
nature of glycan moieties on proteins expressed in higher organisms67. activity relationships. Future work in this area, fuelled by rapid access
Okamoto and co-workers generated the N-glycosylated Ser-CCL1 to mirror-image proteins by chemical synthesis, may help to provide
chemokine (glyco-L-Ser-CCL1) and, using a convergent NCL strategy, vital structural information across a diverse range of targets. This will
constructed non-glycosylated D-Ser-CCL1. Although a quasi-racemic help to improve our understanding of the function of these important
mixture of glyco-L-Ser-CCL1 and D-Ser-CCL1 facilitated the generation biomolecules and how they might be manipulated for therapeutic
of diffraction quality crystals, unfortunately, the complete structure of design or protein engineering.
the glycoprotein could not be determined owing to the inherent dis-
order of the glycan. However, an enantiomeric comparison of the Mirror-image screening technologies
core protein revealed that the glycan moiety did not influence the native In recent years, an increase in our understanding of the molecular basis
protein structure. As disordered glycans have a crucial role in the of disease processes and a desire to tackle targets that may not be drug-
function and stability of proteins, further developments in racemic gable using small molecules have led to the emergence of new classes
crystallographic techniques are required to enable practitioners to of therapeutics, including biologics (a group of therapeutics derived
successfully solve crystal structures of glycoproteins. from natural biomolecules)75. In particular, the exquisite potency,
More recently, the synthetic, monomeric mirror-image protein, target specificity and favourable safety profiles have led to renewed
D-ubiquitin, was used to facilitate the crystallization of oligomeric interests in peptides and proteins as drug leads76. To date, more than
L-ubiquitin dimers, trimers and tetramers via quasi-racemic crystal- 80 peptide drugs have reached the market, and there are now over 100
lography68 (Fig. 2e). Non-covalent assembly led to the co-crystallization FDA-approved monoclonal antibody protein therapies in the clinic77,78.
of the D-ubiquitin monomer with L-oligomers and the formation of However, there are intrinsic limitations of these molecules as drugs,
pseudo mirror images of the oligomers. Using this strategy, crystal including susceptibility to proteolytic degradation by naturally occur-
structures were solved for tri-ubiquitin and tetra-ubiquitin chains at ring enzymes leading to short plasma half-lives, potential immune
resolutions of 1.80 Å and 2.18 Å, respectively. This represented the first activation, and poor oral bioavailability (90% of polypeptide drugs are
report of X-ray crystal structures for ubiquitin oligomers. As such, this delivered by injection)79. Among a range of strategies that have been
novel extension of quasi-racemic crystallography demonstrates that developed (such as PEGylation to increase biological half-life), the use
a single, monomeric D-protein can be used to crystallize recalcitrant of mirror-image peptides and proteins provides an attractive strategy
oligomeric protein structures. for overcoming these shortcomings. More specifically, mirror-image
Investigation into structure–activity relationships of proteins of proteins containing all D-amino acids are less susceptible to protease
therapeutic interest can also be performed using quasi-racemic X-ray degradation as they are not recognized by native proteases and gener-
crystallography. For example, the technique was used to study the ate a minimal immune response, as they are not efficiently processed
effect of replacing native L-Ile and L-threonine (Thr) residues in sea ane- by the immune system, without sacrificing the specificity and affinity
nome stichodactyla (ShK) toxin with their corresponding enantiomers of their L-protein counterparts80,81.
(mirror-image stereoisomers) or diastereomers (non-mirror-image High-throughput screening is a common and powerful tool in drug
stereoisomers)69–72. As a potent inhibitor of the lymphocyte potassium discovery efforts, as large libraries can be rapidly panned against a tar-
channel (Kv1.3), ShK has garnered attention as a potential thera- get of interest, enabling the identification of high-affinity ligands with
peutic for autoimmune diseases73. Ile and Thr are the only common relative ease. Specifically, biomolecular libraries (comprising DNA,
amino acids to have two chiral centres on their side chains, meaning RNA, peptides or proteins) containing millions to trillions of distinct
that to obtain their mirror-image enantiomers, both chiral centres molecules can be subjected to a selection process that enriches only
need to be inverted. If only the α-centre of the L-residue is inverted, those compounds that bind strongly to the target of interest (usually
the diastereomeric amino acids D-allo-Ile or D-allo-Thr are obtained. through replication of the biomolecule of interest). For example, DNA
Initially, it was reported that synthetic D-allo-ShK had a folded libraries can be replicated via polymerase chain reaction (PCR), allowing
structure identical to L-ShK (determined by NMR) and, surprisingly, amplification of the highest affinity ligands for iterative rounds of selec-
potently bound to the Kv1.3 channel69. Subsequent studies by Kent and tion in a process called the Systematic Evolution of Ligands by Expo-
co-workers used quasi-racemic X-ray crystallography to determine nential Enrichment (SELEX) (see the Aptamer selection-reflection and
that D-ShK had the same folding properties as the native L-ShK but was mirror-image SELEX section for details)82. However, peptides and pro-
approximately 50,000-fold less active in blocking the channel70. To fur- teins do not possess a mode of direct replication, meaning that an
ther investigate the impact of inverting Thr and/or Ile side chains, three alternative means is required. Typically, this is achieved by linking the
L-ShK analogues were synthesized, each with a single Ile or Thr residue phenotype (peptide/protein) to its genotype (nucleic acid sequence),
incorporated as its corresponding diastereomer71. Quasi-racemates of which can then be replicated. To date, one of the most common
D-ShK with the three L-ShK analogues provided crystals, which were and fruitful uses of synthetic mirror-image proteins has been their

Nature Reviews Chemistry | Volume 7 | June 2023 | 383–404 390

Review article

deployment in screening campaigns that help to identify novel thera- of amyloid plaques88. These plaques are made up of Aβ peptides, which,
peutic candidates. A description of screening technologies that have in their β-sheet form, form fibrils that can join together89. A synthetic
used synthetic mirror-image proteins is provided in the following D-enantiomer of the amyloid peptide Aβ(1-42) (D-Aβ) was immobilized
sections. and screened against a large randomized peptide library, identifying
one dominant sequence, which had a sub-micromolar dissociation
Mirror-image phage display constant. Subsequently, MIPD was used against D-Aβ to identify another
Phage display is a common screening tool that involves program- novel D-peptide that displayed a moderate binding affinity for L-Aβ
ming a library of filamentous bacteriophages such that each bacte- (IC50 = 7.2 µM)90. Importantly, not only did this peptide inhibit Aβ fibril
riophage expresses a unique peptide on its surface. These peptides formation and redissolve existing fibrils in an in vivo transgenic murine
are screened against a target of interest to find sequences that bind model of AD, but it was also thought to reduce inflammation, redissolve
with high affinity83. The critical link between the genotype of the phage amyloid plaques into monomeric Aβ peptides and facilitate their elimi-
and the observed phenotype (the surface-expressed peptide) is made nation from the brain. Such was the success of this MIPD campaign, the
possible by incorporating the gene that encodes the peptide into the identified peptide has been developed further, including examining its
bacteriophage coat protein gene. Following successful generation of mechanism of action, and additional rational design and optimization,
the phage library, peptides are incubated with the immobilized target ultimately leading to the development of a peptide drug (RD2), which
and those with low affinity are washed away. The phages that bind to completed a phase I clinical trial in 2019 (refs. 91,92). This investigation
the target are eluted and the genetic material is amplified via bacterial is still ongoing but has provided an impressive example of the utility of
infection before repeating the cycle. This leads to enrichment of the MIPD for generating lead compounds with therapeutic potential93–97.
peptides within the pool that exhibit the highest affinity for the target. It is worth noting that Aβ(1–42) is not the only target for which MIPD
In this way, a small number of peptides with the strongest binding can be has been used to search for AD therapeutics; a D-peptide that disrupts
selected from a larger pool of typically >109 distinct peptide sequences. intracellular neurofibrillary tangles (comprised of abnormally folded
The pre-programmed link between genotype and phenotype allows tau proteins) has also been identified98. As these MIPD-generated lead
for the amino acid sequences of the selected binders to be successfully compounds inhibit aggregation of full-length tau protein, they offer
decoded, which then enables them to be chemically synthesized for promising potential not only for the treatment of AD but also for other
further study. The potential for high-throughput screening means ‘tauopathies’ (neurodegenerative diseases caused by tau deposits in
that phage display is an extremely valuable tool for identifying poten- the brain).
tial peptide therapeutics, as evidenced by the fact that FDA-approved There have been several methodological innovations designed to
peptide drugs romiplostim and ecallantide were initially discovered augment the success of D-peptides generated through MIPD for appli-
using this technique79. cations as therapeutics. Recently, Gao and co-workers used bifunc-
In traditional phage display, both the peptide binders and the pro- tional nanoparticles to deliver Aβ(1–42)-specific peptides to the brain,
tein target are comprised entirely of L-amino acids. However, the overcoming issues with blood–brain barrier penetration (which is usu-
limitations of L-peptide therapeutics have resulted in the develop- ally difficult to achieve with peptide drugs)99. Phage display identified
ment of mirror-image phage display (MIPD). Here the D-configured an L-peptide to target the Aβ oligomer and another to target the brain,
mirror-image of the target L-protein is prepared by chemical synthesis which, when incorporated into a bifunctional nanoparticle, resulted
and immobilized, with subsequent phage screening leading to the in a reduction in both Aβ deposition and cognitive deficits. Impor-
identification of high-affinity L-polypeptides (Fig. 3A). Although these tantly, modern drug-delivery strategies such as this may be extended
L-leads would be inactive against the native L-protein, enantiomeric to transport D-peptides identified from MIPD (which possess enhanced
peptides consisting of D-amino acids display mirrored binding inter- stability and reduced immunogenicity) for drug discovery applications.
actions and so can bind the native L-protein with equal affinity. Since Advances in chemical protein synthesis technologies have facili-
being introduced, MIPD has found applications in a diverse range of tated a corresponding leap in the length and complexity of proteins
settings and has contributed significantly to the rising popularity that can be generated as targets for MIPD. For example, NCL has been
of D-peptides and proteins as a new class of stable therapeutics with used for the chemical synthesis of a biotin-tagged D-enantiomer of the
low immunogenicity84–86. 85-residue protein murine double minute 2 homologue (MDM2), an
oncoprotein responsible for the inhibition of tumour suppressor p53
Applications of mirror-image phage display. MIPD was pioneered by (ref. 100). Previous attempts using phage display yielded L-peptides
Kim and co-workers, who synthesized a D-configured c-Src homology 3 that bound to MDM2 (Kd = 3.2 nM); however, mirror-image alternatives
(SH3) domain of Src kinase (D-Src-SH3) via SPPS and used the 60-residue were pursued to try and take advantage of the proteolytic stability of
protein as bait for phage display87. Following selection and synthesis D-peptides101. Screening of N79K-biotin-D-MDM2(25–109) across sev-
of mirror-image variants, a cyclic D-peptide (Pep-D1) that bound to eral MIPD selections afforded L-polypeptides, termed PMIs (p53-MDM2
L-Src-SH3 with a dissociation constant (Kd) of 63 µM was discovered. inhibitors). The corresponding D-enantiomers, D-PMI-α, D-PMI-β and
Interestingly, this D-peptide only partially overlaps with the binding D-PMI-γ, displayed exceptional binding affinity for native MDM2 (219,
site where physiological polyproline-type L-ligands usually bind. As 35 and 53 nM, respectively)102,103 (Fig. 3B). Subsequently, modifications
c-Src activity is essential for osteoclast-mediated bone resorption, were made to improve the poor membrane permeability of the peptides
binders such as Pep-D1 serve as starting points for medicinal chemistry using polymeric micelles and gold nanoparticles, which resulted in a
campaigns targeting osteoporosis. strong increase in cellular p53 signalling104,105. Other modifications
Since the first report of MIPD, this approach has been adopted to improve the properties of these D-PMI peptides have included
to probe novel treatments for a range of pathologies. For exam- their encapsulation in modified liposomes103, targeted mutations
ple, a D-peptide was investigated for treating Alzheimer disease through structural analyses106 and stapling to restrict conformational
(AD), a neurodegenerative disease characterized by an accumulation freedom107.

Nature Reviews Chemistry | Volume 7 | June 2023 | 383–404 391

Review article

A Schematic of the MIPD cycle

D-Peptide
synthesis

Target Target +
D-protein L-protein
Plasmid library
Displayed Synthetic
L-peptide D-peptide
Helper phage Bacteria library

Mirror plane
6 Sequencing and
D-peptide synthesis 1 Phage display
library generation
Purified
5 Phage
plasmids
for analysis amplification

Eluted MIPD cycle DNA library

phage

D-Protein
Phage elution 4 2 Target binding synthesis

Displayed
Wash L-peptide Target
Unbound L-protein
phage
Target
× × × × D-protein
3

Phage separation Mirror plane

B Phage display against MDM2

a O
H
S H
N NH
HN
H N
O O
c
HS O H H
O S
H
D-MDM2 (78–109) OH N NH
H2N HN
H N
77 O NCL O H O
O
O H2N D-MDM2 (25–109) OH
H2N D-MDM2 (25–76) SR

D-PMI-α-L-MDM2 (25–109) complex

MIPD cycle

N79K-biotin-D-MDM2 (25–109) L-PMI-α

C Synthesis and crystal structure of D-VEGF-A

O
HS O
i. NCL H2N D-VEGF-A (1–18) SR
D-VEGF-A (51–102) OH HS HS
H2N 50 ii. Thz HS O
HS O
deprotection Folding
O
D-VEGF-A (20–102) H2N D-VEGF-A (1–102) OH
OH NCL
S O H2N
19
O
D-VEGF-A (20–49) SR
N 19
H
O D-VEGF-A

Nature Reviews Chemistry | Volume 7 | June 2023 | 383–404 392

Review article

Fig. 3 | Mirror-image phage display: a powerful platform for generating purified and the amino acid sequences of the displayed L-peptides are determined.
D-peptide/protein binders against a target of interest. A, Schematic of the The D-peptides of the analogous sequences are then synthesized and screened
mirror-image phage display (MIPD) cycle. (1) Phage display library generation: on the native L-peptide/protein target. B, MIPD against murine double minute 2
plasmid libraries are incorporated to produce a phage library, leading to the homologue (MDM2). a, Schematic for the synthesis of N79K-biotin-D-MDM2
surface expression of L-peptides/proteins. (2) Target binding: the displayed (25–109) protein. b, Representation of the MIPD cycle to produce L-peptide
L-peptides/proteins that have high affinity bind to immobilized D-protein of L-PMI-α with high affinity to the synthetic N79K-biotin-D-MDM2 (25–109)
interest. (3) Phage separation: unbound phages are washed away. (4) Phage protein. c, Co-crystal structure of D-PMI-α (yellow) and L-MDM2 (25–109) (PDB
elution: high-affinity binders are eluted from immobilized D-protein of interest. 3LNJ) (purple). C, Synthesis and crystal structure of D-vascular endothelial
(5) Phage amplification: positive phage binders are amplified via bacterial growth factor (VEGF-A). tert-Butyloxycarbonyl-solid-phase peptide synthesis
infection and the phage cycle is repeated for iterative rounds. (6) Sequencing and constructed three peptide fragments that were assembled via one-pot native
D-peptide synthesis: After multiple cycles, eluted high-affinity phage plasmids are chemical ligation (NCL), yielding D-VEGF-A (PDB 4GLU). PMI, p53-MDM2 inhibitor.

Novel platforms have also been established that enable a virtual with longer sequences or more complex architectures such as trimers
natural product library to be screened against a synthetic mirror-image can lead to enhanced surface interaction with the display library and
target108,109. Of note, this strategy has been used to investigate MDM2 improved target inhibition.
as an anticancer target108. After screening, the identified chiral natural A logical extension of the MIPD platform is to screen a library of
product binders were synthesized in their enantiomeric form to bind to proteins rather than peptides, to maximize the potential for surface
the native protein, exposing a number of natural products that could overlap with the target of interest. Kent and co-workers used this strat-
not have been identified via other screening approaches. In this par- egy when seeking to generate a D-protein agonist for the angiogenic
ticular case, using synthetic D-MDM2(25–109), the authors identified protein vascular endothelial growth factor (VEGF-A), which, when
a novel chiral small-molecule p53-MDM2 inhibitor. overexpressed, can accelerate tumour growth117. D-VEGF-A was syn-
MIPD has also been used to identify D-peptides that can bind to thesized using a one-pot NCL strategy and screened against a phage
other key oncogenic receptors and block immunosuppressive inter- library created through designed mutagenesis of the 56-residue GB1
actions, offering an attractive avenue for the development of novel protein (that is, the B1 domain of streptococcal protein G). The lead
cancer treatments. Examples include T cell immunoglobulin and ITIM D-protein, RFX001, showed reciprocal chiral specificity, binding to
domain110, programmed death ligand 1 (ref. 111) and the epidermal native L-VEGF-A with high affinity (Kd = 85 nM) but, as expected, not
growth factor (EGF)112, all of which were synthesized in their mirror- to D-VEGF-A (Fig. 3C). Racemic crystallography revealed that this
image form and panned against L-peptide libraries to identify target D-protein antagonist covers a large surface area of VEGF-A and blocks
binders with high affinities (Kd values of 5.6, 0.51 and 54 µM, respec- the binding region with its receptor, VEGFR1. To overcome issues with
tively). These promising examples highlight the utility of MIPD for gen- thermal stability, further modifications were made by extending the
erating binders for challenging disease targets and lay the foundation N- and C-termini, leading to an improved analogue (D-RFX037), which
for future work to find improved analogues. was stable at room temperature and had a higher affinity for L-VEGF-A
Innovative approaches have enabled a reduction in the synthetic (Kd = 6 nM)118.
workload required to generate suitable bait for MIPD campaigns. Finally, MIPD has been used to produce heterodimeric bivalent
In 1999, Kim and co-workers113, working to inhibit HIV-1 entry, reported D-proteins. These constructs hold promise as a class of next-generation
the first example of using a binding pocket (rather than a whole pro- therapeutics that may address current limitations in the potency and
tein) to facilitate MIPD screening of peptides. Specifically, the trimeric specificity of smaller D-peptides119. This strategy, again seeking to dis-
peptide IQN17 was used to mimic a trimeric coil of the HIV-1 envelope cover VEGF-A antagonists, involved running MIPD selections against
glycoprotein gp41 that is known to be exposed during viral entry (gp41 D-VEGF-A using two different domain scaffolds (the 53-residue GA and
is responsible for fusion of the viral and cellular membranes). Fmoc- 58-residue Z domains of bacterial proteins G and A). Two proteins were
SPPS was used to afford D-IQN17, which was deployed to competitively identified that bound to VEGF-A (RFX-V1 and RFX-V2, respectively).
screen for binders that might target the binding pocket of full-length Crucially, these proteins were capable of blocking VEGF-A-induced
gp41. Although the first generation of D-peptide binders to L-IQN17 vascular leakages and inhibited tumour growth in vivo, while proving
lacked potency in inhibiting viral entry (confirmed by X-ray crystal- non-immunogenic.
lography), a second generation of D-‘pocket-specific inhibitors of entry’ Phage display is one of the most widely explored applications of
(PIE) were identified using MIPD with a bespoke library and a new gp41 mirror-image peptides and proteins. The examples mentioned earlier
trimer mimic (D-IZN17)114. One peptide, in particular, PIE7 (in its trimeric highlight both the utility and versatility of this methodology for iden-
form), exhibited a 40,000-fold increase in potency (IC50 = 250 pM), tifying novel lead compounds across a broad range of pathologies.
compared with binders from the previous study. Further advances in chemical protein synthesis and phage display
Building on this work, structure-guided MIPD, in which a core technology will continue to drive innovation in this area for the
sequence of the peptide is retained, together with crosslinker optimi- development of new therapeutic candidates in the future.
zation, led to the discovery of a D-peptide trimer (PIE12-trimer), which
displayed a further 100,000-fold increase in binding affinity and an Aptamer selection-reflection and mirror-image SELEX
~80-fold increase in inhibitory potency compared with the second Analogous to the protease sensitivity of native polypeptides, natural
generation of binders115. Furthermore, attachment of a lipid cargo to the single-stranded D-DNA and D-RNA sequences are prone to nuclease
PIE12-trimer to improve pharmacokinetic properties afforded ‘CPT31’, degradation, restricting their potential utility as biopharmaceutical
which entered phase I clinical trials in December 2020 (ref. 116). This and diagnostic agents. Mirror-image oligonucleotides have attracted
work provides a poignant example of how screening for MIPD ‘hits’ considerable attention as a means of overcoming this issue and

Nature Reviews Chemistry | Volume 7 | June 2023 | 383–404 393

Review article

a Selection-reflection cycle b MI-SELEX cycle

ACTGCATGA... ACTGCATGA...
Native D-ssDNA/
RNA library L-ssDNA/RNA library
1 Nucleotide 10 1 Nucleotide 10

6 Cloning and Sequencing- 6

sequencing Library and target Library and target by-synthesis
1 incubation incubation 5
5 1 MI-PCR/RT–PCR
PCR/RT–PCR
amplification Mirror- amplification
Eluted Magnetic Eluted
image
D-ssDNA/RNA bead L-ssDNA/RNA
target
Magnetic Native
bead target
Selection-reflection MI-SELEX cycle

ssDNA/RNA
ssDNA/RNA 2 2 4 elution
elution
4 Target binding
Target binding

Wash Wash

Unbound 3 3 Unbound
D-ssDNA/RNA L-ssDNA/RNA
Magnetic separation Magnetic separation

c Synthesis and utility of D-barnase

D-Aptamer

H2N D-Barnase (1–48) SR O

D-Barnase
NCL
HS O H2N D-Barnase (1–110) NH2

D-Barnase (50–110) NH2

H2N L-Aptamer
49
O
Selection-
reflection
Folding to active
mirror-image
enzyme
L-Barnase
D-Barnase

Fig. 4 | Additional mirror-image screening techniques. a, Selection-reflection: via fluorenylmethoxycarbonyl-solid-phase peptide synthesis and native chemical
a library of D-single-stranded DNA (ssDNA) or RNA is screened against a mirror- ligation (NCL), for use in selection-reflection. D-Aptamers with high binding
image (MI) target of interest. b, MI-Systematic Evolution of Ligands by Exponential affinity were isolated and synthesized in their MI form (L-aptamers) to bind to
Enrichment (SELEX): a library of mirror-image L-ssDNA or RNA is screened against the native L-barnase (PDB 1BNR) (blue). RT–PCR, polymerase chain reaction with
the natural target of interest. c, Schematic showing the synthesis of D-barnase, reverse transcription; R indicates an alkyl thioester.

generating binders with improved stability and excellent target affin- panning a library of native D-configured aptamers against a mirror-
ity. To date, two screening approaches have been developed to yield image target. Following amplification, the D-aptamer with highest
mirror-image oligonucleotides: ‘selection-reflection’ and mirror- affinity for the mirror-image target is synthesized in its L-configuration
image SELEX (MI-SELEX). Both strategies rely on the SELEX process, to bind to the natural target120 (Fig. 4a). By contrast, MI-SELEX screens
whereby a target of interest is immobilized on magnetic beads and a mirror-image L-DNA/RNA library against a natively configured
incubated with a library of single-stranded DNA (ssDNA) or RNA. The target. Mirror-image PCR (MI-PCR), which requires a mirror-image
chains of ssDNA or RNA (aptamers) that are tightly bound to the target D-polymerase enzyme, is followed by sequencing-by-synthesis to
remain, whereas those that weakly bind are washed away. Positive identify the best L-aptamer binders (see the Mirror-image replication
DNA/RNA binders are eluted and amplified via PCR/PCR with reverse and transcription section for the only reported example using this
transcription, and the cycle is repeated with the amplified library. platform)121 (Fig. 4b).
These iterative cycles select a handful of aptamers with the most Fürste and co-workers122 were the first to recognize the potential of
promising binding. selection-reflection, screening a combinatorial D-RNA library against
Following a procedure similar to MIPD and noting that nucleic L-adenosine. After iterative rounds of selection, a 58-base oligonu-
acids are naturally D-configured, aptamer selection-reflection involves cleotide D-RNA ligand was identified that, when synthesized in its

Nature Reviews Chemistry | Volume 7 | June 2023 | 383–404 394

Review article

mirror-image form, bound to natural D-adenosine with an affinity of Joyce and co-workers120 reported the first selection of D-DNA aptamers
1.7 µM. This platform was also used to discover a 38-base oligonucleo- against a functional enzyme, the 110-residue barnase RNase (Fig. 4c).
tide L-RNA binder to L-arginine that had a Kd of 129 µM123. As this tech- In this study, D-barnase was synthesized and incubated with a library of
nique requires the synthesis of a mirror-image target, its application has native D-aptamers. After seven rounds of selection, D-aptamer binders
been limited to relatively small peptides and RNAs124–127. However, there were synthesized in their mirror form: two of these L-aptamers bound to
have been two reported examples of selection-reflection against larger the natural L-barnase with ~100 nM affinity. These examples represent
mirror-image proteins. Yatime and co-workers128 identified a mixed another important application of mirror-image proteins and highlight
L-RNA and DNA aptamer that binds to anaphylatoxin C5a, whereas the immense value and utility of mirror-image nucleic acid polymers.

a Mirror-image central dogma of molecular biology

D-Protein
(L-RNA) D-polymerase assembly
(L-DNA) D-polymerase

L-RNA assembly
L-mRNA

(L-DNA) assembly L-tRNA

D-Ribosome
L-rRNA
(L-RNA) assembly D-Aminoacyl
(L-tRNA) synthetase

b Synthesis of D-ASFV pol X polymerase

HS O O

D-ASFV (107–174) OH i. Acyl hydrazide to thioester H2N D-ASFV (1–85) NHNH 2

H2N ii. NCL
106 O iii. Desulfurization HS O
iv. Acm deprotection
D-ASFV (87–174) OH
H2N i. Acyl hydrazide
AcmS O 86 O to thioester
ii. NCL
D-ASFV (87–105) NHNH2 iii. Protein folding
H2N
86
O

D-ASFV polymerase X

c Synthesis of D-Dpo4 polymerase

HS O
PG PG O
D-Dpo4 (156–202) SR
H2N D-Dpo4 (282–352) NH2 i. Condensation CbzHN
ii. Side-chain 155 O
deprotection

TrtS i. NCL i. Acyl hydrazide

PG PG O ii. Cbz deprotection to thioester
ii. NCL
D-Dpo4 (204–281) OH
BocHN iii. Protein folding
O
203 O O
D-Dpo4 (1–70) SR
N
H NCL
HS O
D-Dpo4-3C
D-Dpo4 (72–154) NHNH2
H2N
71 O

Fig. 5 | The mirror-image central dogma with synthetic schemes of key synthesized by fluorenylmethoxycarbonyl-solid phase peptide synthesis, with
mirror-image D-polymerases, D-African swine fever virus polymerase either N-terminal (cysteine, acetamidomethyl (Acm)-cysteine or free amine) or
X and mutant of Sulfolobus solfataricus D-Dpo4. a, Schematic of mirror- C-terminal (acyl hydrazide or free acid). c, Synthesis of D-Dpo4 polymerase (PDB
image central dogma of molecular biology including mirror-image variants 3PR4). Five peptide fragments, synthesized by fluorenylmethoxycarbonyl-solid
of biomolecules required to assemble a D-protein by mimicking the native phase peptide synthesis, with either N-terminal (cysteine, trityl (Trt)-cysteine with
replication130,132,135,137,143, transcription130,136,139, reverse transcription136 and tert-butyloxycarbonyl (Boc)-protected amine, free amine or acetylated amine) or
translation139,144 machinery of the cell. b, Synthesis of D-African swine fever C-terminal (acyl hydrazide, thioester, free acid or amide). Cbz, carbobenzyloxy;
virus polymerase X (D-ASFV pol X) (PDB 2M2W). Three peptide fragments, NCL, native chemical ligation; PG, protecting group.

Nature Reviews Chemistry | Volume 7 | June 2023 | 383–404 395

Review article

A Synthesis of mutant split-protein mirror-image Pfu DNA polymerase

S O

S O D-PfuN (103–162) NHNH2

N
D-PfuN (164–222) NHNH2 102
N O O
163 CF3
O O
CF3
i. Acyl hydrazide i. Acyl hydrazide to i. Acyl hydrazide
HS O to thioester thioester to thioester
ii. NCL ii. NCL ii. NCL
D-PfuN (224–275) NHNH2 iii. TFA-thiazolidine iii. TFA-thiazolidine iii. Desulfurization
H2N deprotection deprotection
223 O O

His6 D-PfuN (1–39) NHNH2

HS O
i. Acyl hydrazide to thioester
D-PfuN (41–101) NHNH2 ii. NCL
H2N
40
O O
S
O
S D-PfuN (319–367) NHNH2
N
D-PfuN (369–407) NHNH2 318
N O O
368 CF3
O O
CF3
i. Acyl hydrazide i. Acyl hydrazide i. Acyl hydrazide
to thioester to thioester to thioester i. Acyl hydrazide to thioester
HS O ii. NCL
ii. NCL ii. NCL ii. NCL
iii. Desulfurization
D-PfuN (409–467) OH iii. TFA-thiazolidine iii. TFA-thiazolidine iii. TFA-thiazolidine
H2N iv. Acm deprotection
deprotection deprotection deprotection
408 O S O

D-PfuN (277–317) NHNH2

N
276
O O
CF3

O
S
D-PfuC (541–595) NHNH2
N i. Acyl hydrazide to thioester
O 540
O ii. NCL
O
CF3 iii. Desulfurization
H2N D-PfuC (468–500) NHNH2
iv. Acm deprotection

HS O i. Acyl hydrazide to thioester i. Acyl hydrazide to thioester

ii. NCL ii. NCL
D-PfuC (502–539) NHNH2
H2N
501 O S O

D-PfuC (597–651) NHNH2

HS O N
596
D-PfuC (716–775) OH O O
H2N CF3
715 O
i. Acyl hydrazide to thioester i. Acyl hydrazide to thioester
HS O ii. NCL ii. NCL
iii. Desulfurization iii. Desulfurization
D-PfuC (653–714) NHNH2
H2N
652 O

B Two-step assembly of 1.5 kb gene

a b
~90 bp
1 Denaturation Overlapping
primer sequence 1 Denaturation
2 Annealing
2 Annealing
3 Elongation
3 Elongation
Flanking 90 bp
oligonucleotide ~500 bp Oligonucleotide
DNA polymerase
Overlapping sequence

Flanking 90 bp
oligonucleotide

~1.5 kb
~500 bp

Nature Reviews Chemistry | Volume 7 | June 2023 | 383–404 396

Review article

Fig. 6 | Construction of critical componentry required for the generation of two-step procedure. a, Small 90 bp oligonucleotides with overlapping sequences
mirror-image life. A, Synthesis of mutant split-protein mirror-image Pyrococcus (that behave like conventional primers) are extended through multiple iterations
furiosus (Pfu) DNA polymerase (PDB 3A2F). Fifteen peptide fragments, of polymerase chain reaction (PCR) to form an approximately 500-bp-long
synthesized by fluorenylmethoxycarbonyl-solid-phase peptide synthesis, with oligonucleotide. b, Three approximately 500-bp-long oligonucleotides that
either N-terminal; cysteine, trifluoroacetic acid (TFA)-thiazolidine or free amine, also have overlapping sequences along with 90 bp oligonucleotides as primers
or C-terminal; acyl hydrazide or free acid and internal acetamidomethyl (Acm)- undergo PCR to assemble the full-length nucleotide (approximately 1.5 kb).
protected cysteine (not shown in figure). B, Assembly of the 1.5 kb gene via a His6, polyhistidine tag; NCL, native chemical ligation.

Mirror-image life the functional activity of these constructs, the same strategy could be
The creation of mirror-image life is one of the ultimate applications applied to D-Dpo4-3C mutants to generate a mirror-image polymerase
of synthetic mirror-image proteins. Achieving this grand challenge capable of acting on L-RNA and L-DNA.
requires the assembly of mirror-image copies of the systems that ena- Zhu and co-workers took an alternate approach, initially syn-
ble replication, transcription and translation of genetic information, thesizing mutant L-Dpo4-5m from nine peptide fragments using an
delivering a ‘mirror-image central dogma’ of molecular biology129,130 NCL-desulfurization strategy with hydrazide thioester surrogates133.
(Fig. 5a). This is a daunting and complex synthetic endeavour, involving Subsequent research led to an improved synthetic strategy that
construction of numerous mirror-image components, such as DNA and was applied to the synthesis of D-Dpo4-5m135. Amplification of the
RNA polymerases, deoxyribonuclease triphosphates (dNTPs), ribonu- mirror-image E. coli 5S ribosomal RNA gene was accomplished using
clease triphosphates (NTPs) and aminoacyl-tRNA synthetases (aaRSs). D-Dpo4-5m, demonstrating compatibility with the MI-SELEX process.
However, substantial progress has been made towards developing the However, the replication error rate of this polymerase was high (10−4,
requisite mirror-image biomolecules and basic cellular processes that corresponding to one error in every 10,000 bp), with the authors not-
are critical for creating mirror-image life. ing the low quality of commercially available L-dNTPs as a contributing
factor. Building on this work, a system capable of both mirror-image
Mirror-image replication and transcription transcription and reverse transcription was recently reported136. Spe-
Mirror-image L-nucleic acid replication is reliant on the chemical cifically, synthetic mutants of D-Dpo4-5m were generated possessing
synthesis of D-polymerase enzymes. In a seminal study, Zhu and RNA-dependent DNA polymerase activity. By using an ssDNA template
co-workers130 set out to synthesize the mirror-image of the 174-residue coding for the E. coli 5S ribosomal RNA gene and an L-RNA primer, L-5S
African swine fever virus polymerase X (ASFV pol X), the smallest known ribosomal RNA could be transcribed and then, using an L-DNA primer,
DNA polymerase. The synthetic route used iterative NCL reactions to reverse-transcribed with subsequent amplification by MI-PCR.
condense three peptide fragments, affording D-ASFV pol X (Fig. 5b). The The synthetic D-Dpo4-5m was deployed in the first application of
polymerase activity of D-ASFV pol X was tested using an L-DNA template, MI-SELEX, enabling the discovery of an L-DNA aptamer that bound to
a 12-nucleotide (nt) long L-DNA primer, and L-dNTPs in achiral buffer human thrombin with high affinity (Kd = 22 nM)121. A large, randomized
and subsequently compared with a natural L-ASFV pol X system. For L-DNA library (~1014 sequences) was screened against immobilized
both enzymes, the primers were extended by 6 nt over 4 h; however, native human α-thrombin, and the bound L-DNA strands were eluted
the efficiency was lower for the mirror-image system. This was likely and amplified by MI-PCR, using the synthetic D-Dpo4-5m polymerase.
a result of the mirror-image dNTPs being of lower purity. Reciprocal After iterative rounds of selection, the enriched pool was isolated
chiral specificity was demonstrated whereby this D-polymerase only for L-DNA sequencing-by-synthesis. Despite this achievement, it is
polymerized L-dNTPs and, impressively, D-ASFV pol X could also be important to note that this enzyme is limited by its error-prone nature,
used for L-RNA transcription. By contrast, substitution of D-ASFV pol preventing preparation of longer sequences (>200 nt). However, the
X for native L-ASFV pol X stalled the reaction131. The ability of this ‘left- considerable progress on D-Dpo4-5m offers a glimpse into the potential
handed’ polymerase to facilitate both replication and transcription is capabilities of synthetic mirror-image polymerase enzymes.
remarkable and represents a substantial milestone in the pursuit of Building on these studies, Zhu and co-workers reported the syn-
mirror-image life. One drawback of the D-polymerase was that it was thesis of a staggering 775-residue mirror-image Pyrococcus furiosus
not thermostable during PCR, making it impractical for MI-SELEX. Addi- (Pfu) DNA polymerase, which is commonly used in PCR laboratories
tionally, although smaller D-polymerases may be more synthetically owing to its hyperthermostability and high fidelity137 (Fig. 6A). The
accessible, they have reduced fidelity and polymerase activity. synthesis was accomplished using a split-protein approach, whereby
Following this initial study, the chemical synthesis of a thermo- two fragments (a 467-residue Pfu N-fragment and 308-residue Pfu
stable 352-residue Dpo4 polymerase, a mutant of Sulfolobus solfa- C-fragment) were independently synthesized in parallel and co-folded
taricus was reported by two independent groups132,133. Klussman and in vitro. This mirror-image polymerase accurately assembled the 1.5-kb
co-workers used Fmoc-SPPS and hydrazide-assisted NCL to generate mirror-image 16S ribosomal RNA gene via a two-step procedure138.
a mutant enzyme D-Dpo4-3C bearing three Ala to Cys substitutions (to Three large DNA blocks (~550 nt each) were assembled from shorter
avoid the need for desulfurization) with minimal disruption to the struc- oligonucleotides (~90 nt) by overlapping sequences via PCR, which
ture or function of the enzyme132 (Fig. 5c). L-Dpo4-3C was recombinantly behave as primers, where flanking oligonucleotides were used in excess
expressed for comparison, revealing similar polymerase activity when (Fig. 6B). The three large DNA blocks were then combined through
tested using PCR with templates, primers, and dNTPs of the appropriate a similar assembly from overlapping sequences. Purification using
handedness. Finally, further L-Dpo4-3C mutants were expressed with denaturing PAGE determined that 90% of the final sequences were
additional point mutations that are known to provide Dpo4 with the correct and could be templated for future transcription. In the same
ability to elongate both D-RNA and D-DNA134. In addition to confirming report, Zhu and co-workers also exploited the capacity of L-DNA to

Nature Reviews Chemistry | Volume 7 | June 2023 | 383–404 397

Review article

a Synthesis of D-T7 RNA polymerase

HS O

O D-T7N (71–139) NHNH2

H2N
His6 D-T7N (1–37) NHNH2 70 O

HS i. Acyl hydrazide i. Acyl hydrazide i. Acyl hydrazide

O
to thioester to thioester to thioester
D-T7N (39–69) NHNH2 ii. NCL ii. NCL ii. NCL
H2N iii. Desulfurization
38 O HS O

D-T7N (141–215) NHNH2

H 2N
140 O
AcmS O
HS O
D-T7N (217–259) NHNH2
D-T7N (296–363) H2N
OH
H 2N 216 O
295
O
S O i. Acyl hydrazide i. Acyl hydrazide to thioester
to thioester ii. NCL
D-T7N (261–294) NHNH2 i. Acyl hydrazide to thioester O
N ii. NCL iii. Desulfurization
ii. NCL
260 iii. TFA-thiazolidine iii. Acm deprotection
O O deprotection H2N D-T7C (602–673) NHNH2
CF3

i. Acyl hydrazide HS O
to thioester
i. Acyl hydrazide
ii. NCL D-T7C (675–722) NHNH2
to thioester H2N
ii. NCL iii. Desulfurization
Mirror-image T7 674
RNA polymerase O

AcmS O

D-T7C (724–797) NHNH2

H2N
723
O
HS O
i. Acyl hydrazide
O i. Acyl hydrazide
D-T7M (448–490) NHNH2 to thioester
H 2N to thioester ii. NCL
H2N D-T7M (364–408) NHNH2
447 O ii. NCL iii. Desulfurization HS O
HS O iii. Desulfurization iv. Acm deprotection
iv. Acm deprotection
i. Acyl hydrazide i. Acyl hydrazide D-T7C (844–883) OH
D-T7M (410–446) NHNH2 to thioester to thioester H2N
H2N
ii. NCL ii. NCL 843 O
409 O i. Acyl hydrazide O
S
to thioester
D-T7C (799–842) NHNH2
ii. NCL N
S O iii. Desulfurization 798
iv. TFA-thiazolidine O O
CF3
O D-T7M (492–534) NHNH2 deprotection
S N
491
D-T7M (536–557) NHNH2 O O
N CF3
535
O O
CF3 i. Acyl hydrazide i. Acyl hydrazide
HS O to thioester to thioester
ii. NCL ii. NCL
D-T7M (559–601) OH iii. TFA-thiazolidine iii. TFA-thiazolidine
H2N
558 deprotection deprotection
O

b Synthesis of D-ligase O HS
O
R D-LigA (98–136) N
O H2N H
97 O NH2
H2N D-LigA (17–65) N
H
NH2

i. Acyl derivative i. Acyl derivative

HS to thioester to thioester
O i. Acyl derivative
ii. NCL ii. NCL
D-LigA (67–96) iii. Desulfurization to thioester
N
H 2N H ii. NCL
66 O NH2 iii. Protein folding

S O O

D-LigA (138–195) N
O O N R′
S H
137 O N
D-LigA (197–216) N O H
N NH2
H L-DNA D-ligase (17–268)
N
196 O O H i. NCL i. NCL
ii. Thiazolidine ii. Thiazolidine
HS deprotection deprotection

D-LigA (218–268) G3H6 NH2

H2N
217 O

Nature Reviews Chemistry | Volume 7 | June 2023 | 383–404 398

Review article

Fig. 7 | Synthesis of mirror-image mutant split-protein T7 RNA polymerase by a combination of tert-butyloxycarbonyl in situ neutralization SPPS and
and mirror-image DNA ligase. a, Synthesis of D-T7 RNA polymerase (PDB 1CEZ). fluorenylmethoxycarbonyl-SPPS, with thiazolidine or free amine on the
Eighteen peptide fragments, synthesized by fluorenylmethoxycarbonyl-solid- N-terminus; acyl 3,4-diaminobenzoyl, acyl orthophenylene diamine, activated
phase peptide synthesis (SPPS), with trifluoroacetic acid (TFA)-thiazolidine, acyl benzimidazolinone or with a C-terminal histidine tag. A 16-residue truncation
acetamidomethyl (Acm)-protected cysteine or free amine on the N-terminus; at the N-terminus also served to improve the ease of synthesis without impairing
acyl hydrazide or free acid on the C-terminus and internal Acm-protected ligase function. C, C-terminal protein fragment; M, middle protein fragment;
cysteine residues (not shown). b, Synthesis of Haemophilus influenzae D-ligase N, N-terminal protein fragment; NCL, native chemical ligation; R, lysine (Lys)-NH2;
(D-LigA) (adapted from PDB 2Q2T). Six peptide fragments, synthesized Rʹ, Lys-Lys-NH2.

encode information for bioorthogonal storage. A passage from the Mirror-image translation
manuscript in which Louis Pasteur first mentioned that mirror-image A critical first step towards developing a mirror-image translation
life was encoded into DNA sequences and archived into 11 double- system is the synthesis and assembly of large ribosomal proteins
stranded DNA strands of 220 nt each. These L-DNA strands were stable (r-proteins) and ribosomal RNA subunits that make up the transla-
and amplifiable even after DNase I treatment and were successfully tional machinery. A notable example was the syntheses of three E. coli
sequenced by the mutant D-Dpo4-5m (described earlier). r-proteins (L5, L18 and L25) and mirror-image 5S ribosomal RNA144.
More recently, Xu and Zhu139 reported the synthesis of a 883- To reduce the difficulty of the L18 synthesis, the authors examined
residue D-polymerase, the largest protein prepared by chemical the functional effect of L-L18 phosphorylation at two Ser residues
synthesis to date. This 100 kDa protein, T7 RNA polymerase, is the gold (Ser45 and Ser52). Although no difference in binding to ribosomal
standard enzyme for transcription owing to its excellent efficiency, RNA was observed in the absence of this PTM, phosphorylation at Ser52
fidelity and promoter specificity. Akin to the split-protein design of was maintained, given its potential importance to ribosomal assembly
the Pfu L-DNA D-polymerase, a double-split synthetic strategy was in vivo. After settling on L18S52p as the target mutant, L-L18S52p and
used resulting in three split-protein fragments, with 14 Ile substitu- D-L18S52p were synthesized using iterative NCL desulfurization of
tions (5 leucine (Leu), 1 methionine and 8 valine (Val)) to reduce the three peptide fragments. D-L18S52p was folded and assembled with
cost of the synthesis (Fig. 7a). This design was validated by a compari- transcribed L-5S ribosomal RNA to form a ribonucleoprotein. A similar
son of the recombinant L-wild type and the recombinant double-split synthetic strategy using NCL was applied to assemble D-L5 and D-L25,
L-mutant showing similar transcription efficiencies of a D-DNA tem- ultimately enabling construction of their mirror-image ribonucleo-
plate. Following the total chemical synthesis of the double-split T7 RNA protein complexes. Importantly, the three D-r-proteins were able to
polymerase mutants in both L-forms and D-forms, each enantiomeric simultaneously bind to the mirror-image 5S ribosomal RNA, just as
set of protein fragments was folded in vitro. Transcription activity the natural r-proteins do in the native machinery.
of the mirror-image T7 RNA polymerase was tested using double- An alternative approach to assembling synthetic mirror-image
stranded L-DNA templates coding for Thermus thermophilus 5S, 16S ribosomal machinery involves using the natural ribosome system with
and 23S ribosomal RNAs assembled by MI-PCR. The authors found reprogramming to incorporate D-amino acids into proteins145. There
the enzyme could transcribe the full-length mirror-image 5S, 16S and are substantial ‘building block’ requirements for this process, such as
2.9-kb 23S ribosomal RNAs. Other applications of this mirror-image modified tRNAs and aaRSs that will accept the non-canonical amino
RNA polymerase described in the report included the transcription of acid. Importantly, compatibility with elongation factor-Tu (EF-Tu) is
a polymerase L-ribozyme and a mirror-image riboswitch sensor, both required to deliver the amino-acid-charged tRNA to the ribosome. For
of which were resistant to RNase A degradation. This feat of chemical the additional non-canonical amino acids to be translated, the genetic
protein synthesis represents a major milestone towards accessing code must be expanded so that the codons can be accommodated.
the functional mirror-image proteins necessary for mirror-image However, incorporating consecutive sequences of D-amino acids is
replication, transcription and translation systems. challenging146,147. Suga and co-workers148 reported the translation
Extending past mirror-image DNA polymerases, Sczepanski and of 10 consecutive D-amino acids by fine-tuning the concentration of
Joyce140 have developed a cross-chiral RNA polymerase ribozyme EF-Tu and engineering the tRNA body sequence. By using the flex-
that catalyses mirror-image RNA nucleotides on a mirror-image ible in vitro translation system, a D-aminoacyl-tRNA, tRNAGluE2, was
DNA template to assemble its enantiomer. The enantiomer of the engineered to incorporate D-amino acids using tRNA-aminoacylating
template-dependent ribozyme could polymerize natural NTPs on a ribozymes (flexizymes)149. Altering the EF-Tu concentrations enabled
natural template, assembling the original ribozyme. Each enantiomer an efficient initiation and translocation process, yielding the synthesis
displayed chiral specificity, as RNA polymerization of an activated of a 10-residue D-Ser peptide sequence.
nucleotide with the incorrect chirality would inhibit the process. Recently, Zhu and co-workers150 demonstrated the translation
Building on this work, additional optimization of the polymerase of several protein enzymes without the need for aaRSs. By using only
ribozyme141 and the successful cross-chiral amplification of L-RNA142 flexizyme-charged tRNAs to translate these proteins, common issues
have also been reported. associated with flexizymes (such as low translational yields) could be
Hoheisel and co-workers143 have also made an important addition overcome. The low yields are partly attributable to the lack of recycling
to the mirror-image replication and transcription toolkit by synthesiz- that comes from flexizyme-charging of tRNAs without the addition of
ing a functional 252-residue Haemophilus influenzae L-DNA D-ligase aaRSs151. By concentrating the flexizyme-charged tRNAs and reducing
(D-LigA) (Fig. 7b). Such ligases can be used for the production of long cation (Mg2+) contamination, enzymes including active tryptophan
stretches of DNA. Addition of L-ATP and synthetic D-LigA to a mixture aaRS (TrpRS) and luciferase could be translated. Additionally, mirror-
of L-oligonucleotides resulted in successful oligonucleotide ligation image L-tRNAs could be charged with D-amino acids by an L-flexizyme.
to form longer nucleic acid fragments. The use of mirror-image flexizyme-charged tRNAs represents an

Nature Reviews Chemistry | Volume 7 | June 2023 | 383–404 399

Review article

a ‘Top-down’ and ‘bottom-up’ approaches for constructing a mirror- integrate these systems into a mirror-image cell. There are two schools
image artificial cell
of thoughts about how synthetic cells can be generated — either using
a ‘top-down’ or ‘bottom-up’ approach (Fig. 8a). The top-down strat-
egy requires step-by-step alteration of the components of a living cell
until the bare backbone of a functional cell remains152. However, all the
Top-down components for each step need to be synthesized before they can be
approach
removed or judged to be essential, thus limiting the feasibility of this
approach for constructing mirror-image cells.
The alternative bottom-up approach involves synthesizing and
combining the fewest number of components that would be required
to generate an artificial cell153. As the shape of a cell is crucial for its
function, Dekker and co-workers developed a microfluidic platform
Normal cell Mirror-image cell
to manipulate the morphology of artificial cells to mimic those of
natural cells154 (Fig. 8b). Artificial cells were produced by flowing an
inner aqueous phase through a lipid phase, forming spheres which
could be deformed into rods and discs with adjustable dimensions.
By introducing proteins into artificial cells, the authors demonstrated
Bottom-up
approach that three different filamentous protein networks encapsulated into
rod-like containers showed a different assembly within the cell when
compared with a spherical-like container, highlighting the importance
of cell shape and dimension.
An essential property of a living cell is its ability to replicate infor-
mation to make new cells. Additional developments have therefore
Mirror-image Mirror-image cell involved the creation of simple replicating systems, including a bacte-
cellular components rial cell containing only the essential synthetic genes for controlling
the cell155, and the incorporation of a viral DNA replication system into
b Syntheses of artificial cells Feeding synthetic liposome-based cells156. In the latter case, the Φ29 virus DNA
channel
replication machinery amplified a linear DNA template by self-encoded
Outer proteins and replicated DNA served as a template for gene expression.
aqueous (a)
phase This process, deemed an ‘autocatalytic DNA replication cycle’, demon-
Lipid
phase strates the application of a replication system in vitro. However, both
of these examples were performed using native, non-mirror-image
(b)
Inner components. This reinforces the need for further developments before
aqueous the generation of an entire mirror-image cell may be realized.
phase Transfer

Conclusion and outlook

Lipid (c) Interest in mirror-image biomolecules has flourished in recent years,
phase
Outer with growing applications for these molecules in biology and medi-
aqueous cine. As improved tools for chemical protein synthesis have emerged,
phase
the range of mirror-image peptides and proteins that can be accessed
Feeding has greatly expanded, enabling more detailed investigations into the
channel
structure and function of both mirror-image and native biomolecules.
Fig. 8 | Approaches to creating a mirror-image artificial cell. a, Illustration The introduction of racemic and quasi-racemic X-ray crystallog-
of ‘top-down’ and ‘bottom-up’ approaches for constructing a mirror-image raphy has overcome difficulties associated with crystallizing proteins,
artificial cell. The top-down approach (top panel) requires the synthesis of all thus accelerating the determination of peptide and protein 3D struc-
mirror-image components within a cell before integrating them together. The
tures, which can be leveraged for drug design and protein engineering
bottom-up approach requires the synthesis of a minimal number of components
applications. Similarly, mirror-image peptides and proteins, comprised
to generate a functional artificial cell. b, 2D graphical representation of the
entirely of non-canonical D-amino acids, have found utility in novel
syntheses of artificial cells that can be manipulated in size and dimension where
(a) is the spherical cellular structure, (b) is the rod-like structure and (c) is the
therapeutic approaches, owing to their proteolytic stability and low
disc-like structure. immunogenicity. Progress in this area has been supported by the devel-
opment of an array of mirror-image screening technologies that have
allowed for rapid and high-throughput panning across an incredibly
diverse range of targets and that have already identified several clinical
important step towards mirror-image translation by improving the candidates. With new chemical protein synthesis technologies driv-
ease of generating large D-ribosomal proteins. ing the preparation of larger and ever more complex mirror-image
proteins, it is likely that many more therapeutic leads will emerge from
Towards a mirror cell these mirror-image display technologies in the future.
Following the development of suitable mirror-image replication, tran- Moving forward, further improvements in mirror-image biomol-
scription and translation technologies, the main challenge will be to ecule synthesis and the re-engineering of native biological machinery

Nature Reviews Chemistry | Volume 7 | June 2023 | 383–404 400

Review article

will aid in the procurement of the component building blocks that 12. Liu, W. R., Wang, Y.-S. & Wan, W. Synthesis of proteins with defined posttranslational
modifications using the genetic noncanonical amino acid incorporation approach.
are required for constructing mirror-image life. Longstanding limita- Mol. Biosyst. 7, 38–47 (2011).
tions, such as the high cost and low purity of commercially available 13. Zhao, L. & Lu, W. Mirror image proteins. Curr. Opin. Chem. Biol. 22, 56–61 (2014).
L-nucleotide bases, must be overcome to achieve these goals. Novel 14. Abboud, S. A., Cisse, E. H., Doudeau, M., Bénédetti, H. & Aucagne, V. A straightforward
methodology to overcome solubility challenges for N-terminal cysteinyl peptide
approaches such as enzymatic conversion of native monomers to their segments used in native chemical ligation. Chem. Sci. 12, 3194–3201 (2021).
corresponding mirror-image enantiomers provide promise and are 15. Giesler, R. J., Erickson, P. W. & Kay, M. S. Enhancing native chemical ligation for
exciting avenues for future research. Simple strategies including sub- challenging chemical protein syntheses. Curr. Opin. Chem. Biol. 58, 37–44 (2020).
16. Fuchs, O., Trunschke, S., Hanebrink, H., Reimann, M. & Seitz, O. Enabling cysteine‐free
stituting mirror-image monomers (such as those that are prohibitively native chemical ligation at challenging junctions with a ligation auxiliary capable of base
expensive) for structurally similar surrogates are also likely to have key catalysis. Angew. Chem. Int. Ed. 60, 19483–19490 (2021).
roles in progressing the field. Indeed, this practice has already been 17. Mitchell, N. J. et al. Rapid additive-free selenocystine–selenoester peptide ligation.
J. Am. Chem. Soc. 137, 14011–14014 (2015).
demonstrated, with substitution of D-Ile residues for D-Val or D-Leu in 18. Zhang, Y., Xu, C., Lam, H. Y., Lee, C. L. & Li, X. Protein chemical synthesis by serine and
the syntheses of D-Pfu and D-T7, enabling a reduction in the overall cost threonine ligation. Proc. Natl Acad. Sci. USA 110, 6657–6662 (2013).
19. Bang, D., Pentelute, B. L. & Kent, S. B. H. Kinetically controlled ligation for the convergent
and improvements in the ease of synthesis and purification without
chemical synthesis of proteins. Angew. Chem. Int. Ed. 45, 3985–3988 (2006).
compromising the PCR activity of the final enzymes. 20. Hackenberger, C. P. R. & Schwarzer, D. Chemoselective ligation and modification
The recent report of the total chemical synthesis of T7 RNA D-poly- strategies for peptides and proteins. Angew. Chem. Int. Ed. 47, 10030–10074 (2008).
21. Saxon, E., Armstrong, J. I. & Bertozzi, C. R. A ‘traceless’ Staudinger ligation for the
merase, capable of constructing the larger genes required for ribosomal
chemoselective synthesis of amide bonds. Org. Lett. 2, 2141–2143 (2000).
assembly, is an important milestone on the pathway to mirror-image 22. Muir, T. W., Sondhi, D. & Cole, P. A. Expressed protein ligation: a general method for
life139. However, the ribosomal system is highly complex and, although protein engineering. Proc. Natl Acad. Sci. USA 95, 6705–6710 (1998).
23. Gao, X., Ma, Q. & Zhu, H. Distribution, industrial applications, and enzymatic synthesis
a minimal system comprising of r-proteins, tRNAs and ribosomal RNAs
of D-amino acids. Appl. Microbiol. Biotechnol. 99, 3341–3349 (2015).
(excluding 5S rRNA) may be used to perform translation, it is important 24. Zhang, D.-P. et al. Highly selective synthesis of D-amino acids via stereoinversion of
to note that these components are modified in vivo and the use of corresponding counterpart by an in vivo cascade cell factory. Microb. Cell Fact. 20,
11 (2021).
unmodified variants could negatively impact the overall efficiency
25. Milton, R., Milton, S. & Kent, S. Total chemical synthesis of a D-enzyme: the enantiomers
of the translation process157. As such, another important avenue for of HIV-1 protease show reciprocal chiral substrate specificity [corrected]. Science 256,
future investigation is the exploration of effective strategies for intro- 1445–1448 (1992).
This publication marked the landmark synthesis of a mirror-image protein and provided
ducing native modifications to these components of the translational critical experimental evidence demonstrating reciprocal chiral specificity of proteins.
machinery. More broadly, it should also be noted that the creation 26. Weinstock, M. T., Jacobsen, M. T. & Kay, M. S. Synthesis and folding of a mirror-image
and collation of mirror-image cellular components do not neces- enzyme reveals ambidextrous chaperone activity. Proc. Natl Acad. Sci. USA 111,
11679–11684 (2014).
sarily guarantee the formation of a functioning mirror-image cell — This pivotal work pushed the frontier of chemical protein synthesis and uncovered
considerable work is still needed to investigate the key interactions the ambidextrous folding ability of the GroEL–ES chaperone, priming the field for
and/or components that are needed to generate a true mirror-image cell. development of a generalizable strategy for folding mirror-image proteins.
27. Petersen, M. E., Jacobsen, M. T. & Kay, M. S. Synthesis of tumor necrosis factor α for use
Ultimately, although the creation of mirror-image life remains as a mirror-image phage display target. Org. Biomol. Chem. 14, 5298–5303 (2016).
both a highly attractive and intriguing challenge, the application of 28. Feldmann, M. & Maini, R. N. TNF defined as a therapeutic target for rheumatoid arthritis
mirror-image peptides, proteins and other biomolecules also provides and other autoimmune diseases. Nat. Med. 9, 1245–1250 (2003).
29. Levinson, A. M. et al. Total chemical synthesis and folding of all-L and all-D variants
a wealth of other opportunities. Indeed, building on the remarkable of oncogenic KRas(G12V). J. Am. Chem. Soc. 139, 7632–7639 (2017).
advancements that have been made to date, it is likely that further 30. Wilson, R. M. et al. A fascinating journey into history: exploration of the world of isonitriles
developments in the field will be made to fully exploit mirror-image en route to complex amides. Angew. Chem. Int. Ed. 51, 2834–2848 (2012).
31. Noguchi, T. et al. Synthesis of Grb2 SH2 domain proteins for mirror-image screening
biomolecules for applications across both science and medicine. systems. Bioconjug. Chem. 28, 609–619 (2017).
32. Schmidt, N., Abendroth, F., Vázquez, O. & Hantschel, O. Synthesis of the L- and D-SH2
Published online: 12 May 2023 domain of the leukaemia oncogene Bcr-Abl. RSC Chem. Biol. 3, 1008–1012 (2022).
33. De Rosa, L., Di Stasi, R. & D’Andrea, L. D. Total chemical synthesis by native chemical
ligation of the all-D immunoglobulin-like domain 2 of Axl. Tetrahedron 75, 894–905 (2019).
References 34. Callahan, A. J. et al. Single-shot flow synthesis of D-proteins for mirror-image phage
1. Pasteur, L. Mémoire sur la relation qui peut exister entre la forme cristalline et la display. ChemRxiv https://doi.org/10.26434/chemrxiv-2023-x86xp (2023).
composition chimique, et sur la cause de la polarisation rotatoire. C. R. Acad. Sci. 26, 35. Kent, S. B. H. Novel protein science enabled by total chemical synthesis. Protein Sci. 28,
535–538 (1848). 313–328 (2019).
2. Snelders, H. A. M. in van’t Hoff-Le Bel Centennial Vol. 12, Ch. 5, 66–73 (American Chemical 36. Yeung, H. et al. Radiation damage and racemic protein crystallography reveal the unique
Society, 1975). structure of the GASA/snakin protein superfamily. Angew. Chem. Int. Ed. 55, 7930–7933
3. Cahn, R. S. & Ingold, C. K. Specification of configuration about quadricovalent (2016).
asymmetric atoms. J. Chem. Soc. 131, 612–622 (1951). 37. Hung, L.-W., Kohmura, M., Ariyoshi, Y. & Kim, S.-H. Structural differences in D and
4. Rosanoff, M. A. On Fischer’s classification of stereo-isomers. J. Am. Chem. Soc. 28, L-monellin in the crystals of racemic mixture. J. Mol. Biol. 285, 311–321 (1999).
114–121 (1906). 38. Payne, C. D. et al. Solution NMR and racemic crystallography provide insights into a novel
5. Blackmond, D. G. The origin of biological homochirality. Cold Spring Harb. Perspect. Biol. structural class of cyclic plant peptides. RSC Chem. Biol. 2, 1682–1691 (2021).
11, a032540 (2019). 39. Avital-Shmilovici, M. et al. Fully convergent chemical synthesis of ester insulin:
6. Xie, J. & Schultz, P. G. A chemical toolkit for proteins — an expanded genetic code. determination of the high resolution X-ray structure by racemic protein crystallography.
Nat. Rev. Mol. Cell Biol. 7, 775–782 (2006). J. Am. Chem. Soc. 135, 3173–3185 (2013).
7. Rosano, G. L., Morales, E. S. & Ceccarelli, E. A. New tools for recombinant protein 40. Mandal, K., Pentelute, B. L., Tereshko, V., Kossiakoff, A. A. & Kent, S. B. H. X-ray structure
production in Escherichia coli: a 5‐year update. Protein Sci. 28, 1412–1422 (2019). of native scorpion toxin BmBKTx1 by racemic protein crystallography using direct
8. Wright, T. H., Vallée, M. R. J. & Davis, B. G. From chemical mutagenesis to post‐expression methods. J. Am. Chem. Soc. 131, 1362–1363 (2009).
mutagenesis: a 50 year odyssey. Angew. Chem. Int. Ed. 55, 5896–5903 (2016). 41. Mandal, K. et al. Racemic crystallography of synthetic protein enantiomers used to
9. Noren, C. J., Anthony-Cahill, S. J., Griffith, M. C. & Schultz, P. G. A general method for determine the X-ray structure of plectasin by direct methods. Protein Sci. 18, 1146–1154
site-specific incorporation of unnatural amino acids into proteins. Science 244, 182–188 (2009).
(1989). 42. Pentelute, B. L. et al. Total chemical synthesis and X-ray structure of kaliotoxin by
10. Kreil, G. Conversion of L- to D-amino acids: a posttranslational reaction. Science 266, racemic protein crystallography. Chem. Commun. 46, 8174–8176 (2010).
996–997 (1994). 43. Banigan, J. R. et al. Determination of the X-ray structure of the snake venom protein
11. Mor, A., Amiche, M. & Nicolas, P. Enter a new post-translational modification: D-amino omwaprin by total chemical synthesis and racemic protein crystallography. Protein Sci.
acids in gene-encoded peptides. Trends Biochem. Sci. 17, 481–485 (1992). 19, 1840–1849 (2010).

Nature Reviews Chemistry | Volume 7 | June 2023 | 383–404 401

Review article

44. Luisier, S., Avital-Shmilovici, M., Weiss, M. A. & Kent, S. B. H. Total chemical synthesis 76. D’Aloisio, V., Dognini, P., Hutcheon, G. A. & Coxon, C. R. PepTherDia: database and
of human proinsulin. Chem. Commun. 46, 8177–8179 (2010). structural composition analysis of approved peptide therapeutics and diagnostics.
45. Dang, B. et al. Elucidation of the covalent and tertiary structures of biologically active Ts3 Drug Discov. 26, 1409–1419 (2021).
toxin. Angew. Chem. Int. Ed. 55, 8639–8642 (2016). 77. Wang, L. et al. Therapeutic peptides: current applications and future directions.
46. Huang, Y. C. et al. Synthesis of L‐ and D‐ubiquitin by one‐pot ligation and metal‐free Sig. Transduct. Target. Ther. 7, 48 (2022).
desulfurization. Chem. Eur. J. 22, 7623–7628 (2016). 78. Lyu, X. et al. The global landscape of approved antibody therapies. Antib. Ther. 5, 233–257
47. Teng, P. et al. The folding propensity of α/sulfono-γ-AA peptidic foldamers with both (2022).
left- and right-handedness. Commun. Chem. 4, 58 (2021). 79. Muttenthaler, M., King, G. F., Adams, D. J. & Alewood, P. F. Trends in peptide drug discovery.
48. Yeates, T. O. & Kent, S. B. H. Racemic protein crystallography. Annu. Rev. Biophys. 41, Nat. Rev. Drug Discov. 20, 309–325 (2021).
41–61 (2012). 80. Poduslo, J. F., Curran, G. L., Kumar, A., Frangione, B. & Soto, C. Beta-sheet breaker peptide
49. Mackay, A. L. Crystal enigma. Nature 342, 133 (1989). inhibitor of Alzheimer’s amyloidogenesis with increased blood–brain barrier permeability
50. Zawadzke, L. E. & Berg, J. M. The structure of a centrosymmetric protein crystal. and resistance to proteolytic degradation in plasma. J. Neurobiol. 39, 371–382 (1999).
Proteins Struct. Funct. Genet. 16, 301–305 (1993). 81. Dintzis, H. M., Symer, D. E., Dintzis, R. Z., Zawadzke, L. E. & Berg, J. M. A comparison of the
51. Bunker, R. D. et al. A functional role of Rv1738 in Mycobacterium tuberculosis persistence immunogenicity of a pair of enantiomeric proteins. Proteins 16, 306–308 (1993).
suggested by racemic protein crystallography. Proc. Natl Acad. Sci. USA 112, 4310–4315 82. Darmostuk, M., Rimpelova, S., Gbelcova, H. & Ruml, T. Current approaches in SELEX:
(2015). an update to aptamer selection technology. Biotechnol. Adv. 33, 1141–1161 (2015).
This example showcases the power of racemic protein X-ray crystallography for 83. Scott, J. & Smith, G. Searching for peptide ligands with an epitope library. Science 249,
revealing important structural and functional information from targets that were 386–390 (1990).
previously intractable by standard techniques. 84. Guo, Q.-Y. et al. Channel activity of mirror-image M2 proton channel of influenza A virus is
52. Voskuil, M. I. et al. Inhibition of respiration by nitric oxide induces a Mycobacterium blocked by achiral or chiral inhibitors. Protein Cell 10, 211–216 (2019).
tuberculosis dormancy program. J. Exp. Med. 198, 705–713 (2003). 85. Adaligil, E., Patil, K., Rodenstein, M. & Kumar, K. Discovery of peptide antibiotics composed
53. Gupta, V. K., Kumar, M. M., Singh, D., Bisht, D. & Sharma, S. Drug targets in dormant of D-amino acids. ACS Chem. Biol. 14, 1498–1506 (2019).
Mycobacterium tuberculosis: can the conquest against tuberculosis become a reality? 86. Zhou, J., Du, X., Li, J., Yamagata, N. & Xu, B. Taurine boosts cellular uptake of small
Infect. Dis. 50, 81–94 (2018). D-peptides for enzyme-instructed intracellular molecular self-assembly. J. Am. Chem.
54. Chen, C.-C. et al. Racemic X-ray structure of L-type calcium channel antagonist Soc. 137, 10040–10043 (2015).
Calciseptine prepared by total chemical synthesis. Sci. China Chem. 61, 702–707 (2018). 87. Schumacher, T. N. et al. Identification of D-peptide ligands through mirror-image phage
55. Nandakumar, K. S. et al. A recombinant vaccine effectively induces C5a-specific display. Science 271, 1854–1857 (1996).
neutralizing antibodies and prevents arthritis. PLoS ONE 5, e13511 (2010). This seminal work established mirror-image phage display as a powerful technology
56. Zuo, C. et al. Chemical synthesis and racemic crystallization of rat C5a-desArg. for generating high-affinity D-configured peptide ligands of native L-proteins.
Chin. Chem. Lett. 31, 693–696 (2020). 88. Wiesehan, K. et al. Selection of D-amino-acid peptides that bind to Alzheimer’s disease
57. Cook, W. J., Galakatos, N., Boyar, W. C., Walter, R. L. & Ealick, S. E. Structure of human amyloid peptide Aβ142 by mirror image phage display. ChemBioChem 4, 748–753 (2003).
desArg-C5a. Acta Crystallogr. D 66, 190–197 (2010). 89. Fish, P. V., Steadman, D., Bayle, E. D. & Whiting, P. New approaches for the treatment
58. Wang, C. K. et al. Mirror images of antimicrobial peptides provide reflections on their of Alzheimer’s disease. Bioorg. Med. Chem. Lett. 29, 125–133 (2019).
functions and amyloidogenic properties. J. Am. Chem. Soc. 138, 5706–5713 (2016). 90. Van Groen, T. et al. Reduction of Alzheimer’s disease amyloid plaque load in transgenic
59. Kurgan, K. W. et al. Retention of native quaternary structure in racemic melittin crystals. mice by D3, a D-enantiomeric peptide identified by mirror image phage display.
J. Am. Chem. Soc. 141, 7704–7708 (2019). ChemMedChem 3, 1848–1852 (2008).
This article highlights the potential utility of racemic crystallography for the 91. Bocharov, E. V. et al. All-D-enantiomeric peptide D3 designed for Alzheimer’s disease
characterization of native polypeptide quaternary structures. treatment dynamically interacts with membrane-bound amyloid-β precursors. J. Med.
60. Anderson, D., Terwilliger, T. C., Wickner, W. & Eisenberg, D. Melittin forms crystals which Chem. 64, 16464–16479 (2021).
are suitable for high resolution X-ray structural analysis and which reveal a molecular 92. Kutzsche, J. et al. Safety and pharmacokinetics of the orally available antiprionic
2-fold axis of symmetry. J. Biol. Chem. 255, 2578–2582 (1980). compound PRI-002: a single and multiple ascending dose phase I study. Alzheimers
61. Yan, B., Ye, L., Xu, W. & Liu, L. Recent advances in racemic protein crystallography. Dement. 6, e12001 (2020).
Bioorg. Med. Chem. 25, 4953–4965 (2017). 93. Funke, S. A. et al. Oral treatment with the D-enantiomeric peptide D3 improves the
62. Kent, S. B. Racemic & quasi-racemic protein crystallography enabled by chemical protein pathology and behavior of Alzheimer’s disease transgenic mice. ACS Chem. Neurosci. 1,
synthesis. Curr. Opin. Chem. Biol. 46, 1–9 (2018). 639–648 (2010).
63. Qu, Q. et al. Synthesis of disulfide surrogate peptides incorporating large‐span surrogate 94. Jiang, N. et al. Preclinical pharmacokinetic studies of the tritium labelled D-enantiomeric
bridges through a native‐chemical‐ligation‐assisted diaminodiacid strategy. Angew. Chem. peptide D3 developed for the treatment of Alzheimer´s disease. PLoS ONE 10, e0128553
Int. Ed. 59, 6037–6045 (2020). (2015).
64. Pentelute, B. L., Gates, Z. P., Dashnau, J. L., Vanderkooi, J. M. & Kent, S. B. H. Mirror image 95. Leithold, L. H. E. et al. Pharmacokinetic properties of a novel D-peptide developed to
forms of snow flea antifreeze protein prepared by total chemical synthesis have identical be therapeutically active against toxic β-amyloid oligomers. Pharm. Res. 33, 328–336
antifreeze activities. J. Am. Chem. Soc. 130, 9702–9707 (2008). (2016).
65. Pentelute, B. L. et al. X-ray structure of snow flea antifreeze protein determined by racemic 96. Van Groen, T. et al. The Aβ oligomer eliminating D-enantiomeric peptide RD2 improves
crystallization of synthetic protein enantiomers. J. Am. Chem. Soc. 130, 9695–9701 (2008). cognition without changing plaque pathology. Sci. Rep. 7, 16275 (2017).
66. Okamoto, R. et al. (Quasi-)racemic X-ray structures of glycosylated and non-glycosylated 97. Schemmert, S. et al. In vitro and in vivo efficacies of the linear and the cyclic version of an
forms of the chemokine Ser-CCL1 prepared by total chemical synthesis. Angew. Chem. all-D-enantiomeric peptide developed for the treatment of Alzheimer’s disease. Int. J.
Int. Ed. 53, 5194–5198 (2014). Mol. Sci. 22, 6553 (2021).
67. Dwek, R. A. Glycobiology: toward understanding the function of sugars. Chem. Rev. 96, 98. Malhis, M. et al. Potent Tau aggregation inhibitor D-peptides selected against Tau-repeat
683–720 (1996). 2 using mirror image phage display. ChemBioChem 22, 3049–3059 (2021).
68. Gao, S. et al. Monomer/oligomer quasi-racemic protein crystallography. J. Am. Chem. Soc. 99. Zhang, X. et al. Brain targeting and Aβ binding bifunctional nanoparticles inhibit amyloid
138, 14497–14502 (2016). protein aggregation in APP/PS1 transgenic mice. ACS Chem. Neurosci. 12, 2110–2121
This work describes the application of quasi-racemic protein crystallography (2021).
to facilitate the first successful crystallization of ubiquitin oligomers. 100. Brown, C. J., Lain, S., Verma, C. S., Fersht, A. R. & Lane, D. P. Awakening guardian angels:
69. Beeton, C. et al. The D-diastereomer of ShK toxin selectively blocks voltage-gated K+ drugging the p53 pathway. Nat. Rev. Cancer 9, 862–873 (2009).
channels and inhibits T lymphocyte proliferation. J. Biol. Chem. 283, 988–997 (2008). 101. Pazgier, M. et al. Structural basis for high-affinity peptide inhibition of p53 interactions
70. Dang, B., Kubota, T., Mandal, K., Bezanilla, F. & Kent, S. B. H. Native chemical ligation at with MDM2 and MDMX. Proc. Natl Acad. Sci. USA 106, 4665–4670 (2009).
Asx-Cys, Glx-Cys: chemical synthesis and high-resolution X-ray structure of ShK toxin 102. Liu, M. et al. A left-handed solution to peptide inhibition of the p53–MDM2 interaction.
by racemic protein crystallography. J. Am. Chem. Soc. 135, 11911–11919 (2013). Angew. Chem. Int. Ed. 49, 3649–3652 (2010).
71. Dang, B. et al. Inversion of the side-chain stereochemistry of indvidual Thr or Ile residues 103. Liu, M. et al. D-peptide inhibitors of the p53–MDM2 interaction for targeted molecular
in a protein molecule: impact on the folding, stability, and structure of the ShK toxin. therapy of malignant neoplasms. Proc. Natl Acad. Sci. USA 107, 14321–14326 (2010).
Angew. Chem. Int. Ed. 56, 3324–3328 (2017). 104. Chen, X. et al. A stapled peptide antagonist of MDM2 carried by polymeric micelles
72. Dang, B., Chhabra, S., Pennington, M. W., Norton, R. S. & Kent, S. B. H. Reinvestigation sensitizes glioblastoma to temozolomide treatment through p53 activation. J. Control. Rel.
of the biological activity of D-allo-ShK protein. J. Biol. Chem. 292, 12599–12605 (2017). 218, 29–35 (2015).
73. Wulff, H., Castle, N. A. & Pardo, L. A. Voltage-gated potassium channels as therapeutic 105. Bian, Z. et al. Awakening p53 in vivo by D-peptides-functionalized ultra-small nanoparticles:
targets. Nat. Rev. Drug Discov. 8, 982–1001 (2009). overcoming biological barriers to D-peptide drug delivery. Theranostics 8, 5320–5335
74. Kent, S. B. H. Inventing Synthetic Methods to Discover How Enzymes Work. 283–288 (2018).
(GNT-Verlag GmbH, 2022). 106. Zhan, C. et al. An ultrahigh affinity D-peptide antagonist of MDM2. J. Med. Chem. 55,
75. Li, Y., Cao, X. X., Tian, C. L. & Zheng, J. S. Chemical protein synthesis-assisted high- 6237–6241 (2012).
throughput screening strategies for D-peptides in drug discovery. Chin. Chem. Lett. 31, 107. Kannan, S. et al. Macrocyclization of an all-D linear α-helical peptide imparts cellular
2365–2374 (2020). permeability. Chem. Sci. 11, 5577–5591 (2020).

Nature Reviews Chemistry | Volume 7 | June 2023 | 383–404 402

Review article

108. Noguchi, T. et al. Screening of a virtual mirror-image library of natural products. 139. Xu, Y. & Zhu, T. F. Mirror-image T7 transcription of chirally inverted ribosomal and functional
Chem. Commun. 52, 7653–7656 (2016). RNAs. Science 378, 405–412 (2022).
109. Shu, K. et al. Development of mirror-image screening systems for XIAP BIR3 domain The assembly of a functional mirror-image T7 RNA polymerase represents a
inhibitors. Bioconjug. Chem. 30, 1395–1404 (2019). monumental feat of total chemical synthesis and a pivotal breakthrough on the
110. Zhou, X. et al. A novel D‐peptide identified by mirror‐image phage display blocks journey towards mirror-image life.
TIGIT/PVR for cancer immunotherapy. Angew. Chem. Int. Ed. 59, 15114–15118 (2020). 140. Sczepanski, J. T. & Joyce, G. F. A cross-chiral RNA polymerase ribozyme. Nature 515,
111. Chang, H. N. et al. Blocking of the PD‐1/PD‐L1 interaction by a D‐peptide antagonist 440–442 (2014).
for cancer immunotherapy. Angew. Chem. Int. Ed. 54, 11760–11764 (2015). 141. Tjhung, K. F., Sczepanski, J. T., Murtfeldt, E. R. & Joyce, G. F. RNA-catalyzed cross-chiral
112. Díaz‐Perlas, C. et al. Protein chemical synthesis combined with mirror‐image polymerization of RNA. J. Am. Chem. Soc. 142, 15331–15339 (2020).
phage display yields D‐peptide EGF ligands that block the EGF–EGFR interaction. 142. Bare, G. A. L. & Joyce, G. F. Cross-chiral, RNA-catalyzed exponential amplification of RNA.
ChemBioChem 20, 2079–2084 (2019). J. Am. Chem. Soc. 143, 19160–19166 (2021).
113. Eckert, D. M., Malashkevich, V. N., Hong, L. H., Carr, P. A. & Kim, P. S. Inhibiting HIV-1 entry. 143. Weidmann, J., Schnölzer, M., Dawson, P. E. & Hoheisel, J. D. Copying life: synthesis of an
Cell 99, 103–115 (1999). enzymatically active mirror-image DNA-ligase made of D-amino acids. Cell Chem. Biol.
114. Welch, B. D., Vandemark, A. P., Heroux, A., Hill, C. P. & Kay, M. S. Potent D-peptide 26, 645–651 (2019).
inhibitors of HIV-1 entry. Proc. Natl Acad. Sci. USA 104, 16828–16833 (2007). 144. Ling, J. J. et al. Mirror‐image 5S ribonucleoprotein complexes. Angew. Chem. Int. Ed. 59,
115. Welch, B. D. et al. Design of a potent D-peptide HIV-1 entry inhibitor with a strong barrier 3724–3731 (2020).
to resistance. Virol. J. 84, 11235–11244 (2010). 145. Jin, X., Park, O.-J. & Hong, S. H. Incorporation of non-standard amino acids into proteins:
116. Redman, J. S. et al. Pharmacokinetic and chemical synthesis optimization of a potent challenges, recent achievements, and emerging applications. Appl. Microbiol. Biotechnol.
D-peptide HIV entry inhibitor suitable for extended-release delivery. Mol. Pharm. 15, 103, 2947–2958 (2019).
1169–1179 (2018). 146. Li, X. & Liu, C. C. Biological applications of expanded genetic codes. ChemBioChem 15,
117. Mandal, K. et al. Chemical synthesis and X-ray structure of a heterochiral {D-protein 2335–2341 (2014).
antagonist plus vascular endothelial growth factor} protein complex by racemic 147. Katoh, T., Passioura, T. & Suga, H. Advances in in vitro genetic code reprogramming in
crystallography. Proc. Natl Acad. Sci. USA 109, 14779–14784 (2012). 2014–2017. Synth. Biol. 3, ysy008 (2018).
This article expanded the scope of mirror-image phase display by demonstrating the 148. Katoh, T., Tajima, K. & Suga, H. Consecutive elongation of D-amino acids in translation.
value of screening protein (rather than peptide) libraries for generating ligands with Cell Chem. Biol. 24, 46–54 (2017).
excellent selectivity and affinity. 149. Goto, Y., Katoh, T. & Suga, H. Flexizymes for genetic code reprogramming. Nat. Protoc. 6,
118. Uppalapati, M. et al. A potent D-protein antagonist of VEGF-A is nonimmunogenic, 779–790 (2011).
metabolically stable, and longer-circulating in vivo. ACS Chem. Biol. 11, 1058–1065 (2016). 150. Chen, J., Chen, M. & Zhu, T. F. Translating protein enzymes without aminoacyl-tRNA
119. Marinec, P. S. et al. A non-immunogenic bivalent D-protein potently inhibits retinal synthetases. Chem 7, 786–798 (2021).
vascularization and tumor growth. ACS Chem. Biol. 16, 548–556 (2021). 151. Murakami, H., Ohta, A., Ashigai, H. & Suga, H. A highly flexible tRNA acylation method
120. Olea, C. Jr, Weidmann, J., Dawson, P. E. & Joyce, G. F. An L-RNA aptamer that binds and for non-natural polypeptide synthesis. Nat. Methods 3, 357–359 (2006).
inhibits RNase. Chem. Biol. 22, 1437–1441 (2015). 152. Cameron, D. E., Bashor, C. J. & Collins, J. J. A brief history of synthetic biology. Nat. Rev.
121. Chen, J., Chen, M. & Zhu, T. F. Directed evolution and selection of biostable L-DNA Microbiol. 12, 381–390 (2014).
aptamers with a mirror-image DNA polymerase. Nat. Biotechnol. 40, 1601–1609 (2022). 153. Bedau, M. et al. Life after the synthetic cell. Nature 465, 422–424 (2010).
This report discloses an innovative alternative strategy for the generation of mirror- 154. Fanalista, F. et al. Shape and size control of artificial cells for bottom-up biology.
image aptamer binders, relying on the panning of a mirror-image library against a ACS Nano 13, 5439–5450 (2019).
native target of interest and amplification via 155. Gibson, D. G. et al. Creation of a bacterial cell controlled by a chemically synthesized
mirror-image PCR. genome. Science 329, 52–56 (2010).
122. Klußmann, S., Nolte, A., Bald, R., Erdmann, V. A. & Fürste, J. P. Mirror-image RNA that 156. van Nies, P. et al. Self-replication of DNA by its encoded proteins in liposome-based
binds D-adenosine. Nat. Biotechnol. 14, 1112–1115 (1996). synthetic cells. Nat. Commun. 9, 1583 (2018).
These back-to-back publications disclosed the development of selection-reflection 157. Murase, Y., Nakanishi, H., Tsuji, G., Sunami, T. & Ichihashi, N. In vitro evolution of
as a next-generation combinatorial screening tool for the discovery of stable aptamer unmodified 16S rRNA for simple ribosome reconstitution. ACS Synth. Biol. 7, 576–583
ligands. (2018).
123. Nolte, A., Klußmann, S., Bald, R., Erdmann, V. A. & Fürste, J. P. Mirror-design of 158. Mirwaldt, C., Korndorfer, I. & Huber, R. The crystal structure of dihydrodipicolinate
L-oligonucleotide ligands binding to L-arginine. Nat. Biotechnol. 14, 1116–1119 (1996). synthase from Escherichia coli at 2.5 Å resolution. J. Mol. Biol. 246, 227–239 (1995).
These back-to-back publications disclosed the development of selection-reflection 159. Merrifield, R. B. Solid phase peptide synthesis. I. The synthesis of a tetrapeptide. J. Am.
as a next-generation combinatorial screening tool for the discovery of stable aptamer Chem. Soc. 85, 2149–2154 (1963).
ligands. 160. Conibear, A. C., Watson, E. E., Payne, R. J. & Becker, C. F. W. Native chemical ligation in
124. Williams, K. P. et al. Bioactive and nuclease-resistant L-DNA ligand of vasopressin. protein synthesis and semi-synthesis. Chem. Soc. Rev. 47, 9046–9068 (2018).
Proc. Natl Acad. Sci. USA 94, 11285–11290 (1997). 161. Kulkarni, S. S., Sayers, J., Premdjee, B. & Payne, R. J. Rapid and efficient protein synthesis
125. Purschke, W. G., Eulberg, D., Buchner, K., Vonhoff, S. & Klussmann, S. An L-RNA-based through expansion of the native chemical ligation concept. Nat. Rev. Chem. 2, 0122 (2018).
aquaretic agent that inhibits vasopressin in vivo. Proc. Natl Acad. Sci. USA 103, 5173–5178 162. Bondalapati, S., Jbara, M. & Brik, A. Expanding the chemical toolbox for the synthesis
(2006). of large and uniquely modified proteins. Nat. Chem. 8, 407–418 (2016).
126. Sczepanski, J. T. & Joyce, G. F. Binding of a structured D-RNA molecule by an L-RNA 163. Thompson, R. E. & Muir, T. W. Chemoenzymatic semisynthesis of proteins. Chem. Rev.
aptamer. J. Am. Chem. Soc. 135, 13290–13293 (2013). 120, 3051–3126 (2020).
127. Umar, M. I. & Kwok, C. K. Specific suppression of D-RNA G-quadruplex–protein interaction 164. Agouridas, V. et al. Native chemical ligation and extended methods: mechanisms,
with an L-RNA aptamer. Nucleic Acids Res. 48, 10125–10141 (2020). catalysis, scope, and limitations. Chem. Rev. 119, 7328–7443 (2019).
128. Yatime, L. et al. Structural basis for the targeting of complement anaphylatoxin C5a 165. Dawson, P. E., Muir, T. W., Clark-Lewis, I. & Kent, S. B. Synthesis of proteins by native
using a mixed L-RNA/L-DNA aptamer. Nat. Commun. 6, 6481 (2015). chemical ligation. Science 266, 776–779 (1994).
129. Rohden, F., Hoheisel, J. D. & Wieden, H.-J. Through the looking glass: milestones on the The trailblazing development of the native chemical ligation technology underpins
road towards mirroring life. Trends Biochem. Sci. 46, 931–943 (2021). the area of chemical protein synthesis and has enabled construction of larger and
130. Wang, Z., Xu, W., Liu, L. & Zhu, T. F. A synthetic molecular system capable of mirror-image more complex polypeptide architectures.
genetic replication and transcription. Nat. Chem. 8, 698–704 (2016). 166. Wan, Q. & Danishefsky, S. J. Free-radical-based, specific desulfurization of cysteine:
131. Joyce, G. F. et al. Chiral selection in poly(C)-directed synthesis of oligo(G). Nature 310, a powerful advance in the synthesis of polypeptides and glycopolypeptides. Angew.
602–604 (1984). Chem. Int. Ed. 46, 9248–9252 (2007).
132. Pech, A. et al. A thermostable D-polymerase for mirror-image PCR. Nucleic Acids Res. 45, 167. Jin, K., Li, T., Chow, H. Y., Liu, H. & Li, X. P−B desulfurization: an enabling method for
3997–4005 (2017). protein chemical synthesis and site-specific deuteration. Angew. Chem. Int. Ed. 56,
133. Xu, W. et al. Total chemical synthesis of a thermostable enzyme capable of polymerase 14607–14611 (2017).
chain reaction. Cell Discov. 3, 17008 (2017). 168. Chisholm, T. S., Clayton, D., Dowman, L. J., Sayers, J. & Payne, R. J. Native chemical
134. Sherrer, S. M., Beyer, D. C., Xia, C. X., Fowler, J. D. & Suo, Z. Kinetic basis of sugar selection ligation–photodesulfurization in flow. J. Am. Chem. Soc. 140, 9020–9024 (2018).
by a Y-family DNA polymerase from Sulfolobus solfataricus P2. Biochemistry 49, 169. Fang, G.-M. et al. Protein chemical synthesis by ligation of peptide hydrazides.
10179–10186 (2010). Angew. Chem. Int. Ed. 50, 7645–7649 (2011).
135. Jiang, W. et al. Mirror-image polymerase chain reaction. Cell Discov. 3, 17037 (2017). 170. Zheng, J.-S., Tang, S., Qi, Y.-K., Wang, Z.-P. & Liu, L. Chemical synthesis of proteins using
136. Wang, M. et al. Mirror-image gene transcription and reverse transcription. Chem 5, peptide hydrazides as thioester surrogates. Nat. Protoc. 8, 2483–2495 (2013).
848–857 (2019). 171. Flood, D. T. et al. Leveraging the Knorr pyrazole synthesis for the facile generation
137. Fan, C., Deng, Q. & Zhu, T. F. Bioorthogonal information storage in L-DNA with a high-fidelity of thioester surrogates for use in native chemical ligation. Angew. Chem. Int. Ed. 57,
mirror-image Pfu DNA polymerase. Nat. Biotechnol. 39, 1548–1555 (2021). 11634–11639 (2018).
138. Xiong, A. S. et al. A simple, rapid, high-fidelity and cost-effective PCR-based two-step 172. Fang, G.-M., Wang, J.-X. & Liu, L. Convergent chemical synthesis of proteins by ligation
DNA synthesis method for long gene sequences. Nucleic Acids Res. 32, e98 (2004). of peptide hydrazides. Angew. Chem. Int. Ed. 51, 10347–10350 (2012).

Nature Reviews Chemistry | Volume 7 | June 2023 | 383–404 403

Review article

Acknowledgements Additional information

The authors acknowledge the Australian Research Council Centre of Excellence for Innovations Peer review information Nature Reviews Chemistry thanks Wuyuan Lu and the other,
in Peptide and Protein Science (CE200100012) and the NHMRC Investigator Grant APP1174941 anonymous, reviewer(s) for their contribution to the peer review of this work.
(to R.J.P.) for funding and the Research Training Scholarship Program for PhD support (to A.S.M.,
L.K. and J.W.C.M.). Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in
published maps and institutional affiliations.
Author contributions
K.H. researched data for the article and contributed to writing, preparation of figures and Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this
reviewing and editing the manuscript. A.S.M. contributed to writing, preparation of figures and article under a publishing agreement with the author(s) or other rightsholder(s); author self-
reviewing and editing the manuscript. L.K. and J.W.C.M contributed to preparation of the figures archiving of the accepted manuscript version of this article is solely governed by the terms
and revised the manuscript. R.J.P. revised the manuscript and conceived the overall direction of of such publishing agreement and applicable law.
the manuscript. All authors have given approval to the final version of the manuscript.
© Springer Nature Limited 2023
Competing interests
The authors declare no competing interests.

Nature Reviews Chemistry | Volume 7 | June 2023 | 383–404 404