Stephen A.
Hill
*,
Robert
Steinfort
and
Laura
Hartmann
*
Institute of Organic and Macromolecular Chemistry, Heinrich Heine University Düsseldorf, Universitätsstraße 1, 40225 Düsseldorf, Germany. E-mail: stephen.hill@hhu.de; laura.hartmann@hhu.de
First published on 14th July 2021
The synthesis of sequence-defined oligomers and polymers is an important contemporary research topic due to the potential for creating materials with function tailored directly into the primary structure. Sequence-defined materials have so far been applied to biomimicry, data storage and sensing. Heterocyclic building blocks such as thiolactones and epoxides are key building blocks for producing sequence-defined macromolecules using iterative methodologies. In this review we critically evaluate: the current literature on different heterocyclic strategies using solid- or solution-phase methods for the iterative synthesis of sequence-defined polymers; the challenges surrounding these systems; and the potential future direction of heterocycles, so far not used in this field.
Various synthetic strategies have been applied to obtain SDMs,5 however, in this review, we will focus exclusively on methods based on iterative processes. When considering SDM synthesis, methods employing the iterative coupling of BBs on to a growing macromolecular backbone have clear advantages. Ever since R. B. Merrifield reported the first synthesis of a tetrapeptide on a solid-phase (SP) support,6 the inherent benefits of SP and iterative coupling strategies have been realized for polysaccharide,7,8 polypeptide,9 and polynucleotide construction.10 The benefits of polymer chain immobilization include: (1) ease of purification; (2) coupling iteration; and (3) high yield acquisition through excess reagent usage. Despite time and efficiency advantages SP protocols have limitations too: (1) scale-up challenge; (2) optimising new processes i.e., solvents, linker, concentrations etc.; and (3) potential requirement for protecting groups (PGs). Solution-phase strategies circumvent scale limitations but instead introduce purification requirements, which have, in part, been addressed by soluble fluorous, PEG or polystyrene supports.11,12
Many successful SDM syntheses have been based upon the iterative assembly of acyclic BBs. Using tailored, acyclic BBs holds many advantages as the chemist has control over backbone length and composition e.g., tuning hydrophilic or hydrophobic properties; spacing between side-chains, etc. Custom synthesis allows libraries of acyclic BBs to be created: with demarcated lengths, elemental composition, and side- and end-groups, including BBs with or without PGs.13 However, given the contemporary drive for ever increasing efficiencies, non-commercial, tailored acyclics may be disfavoured due to “atom economy” considerations or the time investment often required for the multi-step syntheses.14 Solutions to overcoming inefficiencies, particularly at large scale, are being addressed i.e. we recently published an automation-compatible protocol to recover unreacted BBs.15
Alongside acyclic BB development, cyclic BBs have been developed too. Although not entirely exclusive, (hetero)cyclic-based strategies have been developed as they can offer the following benefits: (1) yielding products with bi- or multi-functionality; (2) proceeding with high regio- and/or stereoselectivity; and (3) being ring-opened with a range of nucleophiles, typically avoiding the need for activation or catalytic agents. Nevertheless, ring openings bring challenges too, namely (un)foreseen cascade/side reactions and product mixtures.
In this review, we shall highlight and discuss strategies and techniques which iteratively construct monodisperse SDMs by employing heterocyclic BBs. The scope of this review will encompass SP and in-solution approaches which specifically employ heterocyclic BBs, that are restricted to cyclic compounds with 3–6 endocyclic atoms, with at least one endocyclic atom being a heteroatom, and the reactivity is directly associated with the cyclic motif's inherent chemistry, and not that of residual groups. Although much literature exists regarding carbocyclic (aliphatic and aromatic) ring synthesis and transformation,16–18 there exists a rich set of diverse heterocycles which offer distinct advantages including, but not limited to: (1) their synthesis/sourcing from renewable resources; and [2] their reactivity generated from endo- and exocyclic polarity differences. Heterocycles which have been reported for SDM synthesis shall be addressed, but also potential future direction and opportunities e.g. of building blocks so far not used within the SDM field shall be presented.
For the purposes of this review, we shall introduce a distinction to differentiate cyclic BBs under the categories of “active” and “latent”. These titles designate the outcome of a given BB's coupling to a macromolecule (Fig. 1). If a BB is “active”, the heterocycle is chemically transformed upon SDM incorporation. “Latent” BBs have an intact, unreacted heterocycle after SDM incorporation, retaining their specific reactivity until a later juncture.
![]() | ||
Fig. 1 Illustration of active and latent strategies of generic heterocycle addition to a growing sequence-defined macromolecule. |
Throughout this review automation shall be highlighted given its importance in offering productivity and safety advantages. Given the topic's importance, automated SP syntheses for biomacromolecules,7,19 and SDMs have already been reviewed.12 It should be noted, larger macrocyclic rings are also known to have been utilized to generate polymeric sequence-control and sequence-definition. Macrocyclic ring opening to generate polymers bring the advantage of rapidly accessing higher molecular weight polymers. For example, Gutekunst and Hawker in 2015 reported the generation of an unstrained, cyclic macromonomer with a defined BB order encoded into the macromonomer ring.20 The macrocycle also contained an enyne trigger which facilitated ring-opening metathesis polymerization after exposure to Grubbs catalyst, producing high molecular weight, sequence-defined polymers. Similar macrocycle strategies to achieve sequence-definition or control are known, but are beyond the scope of this review.21–24
![]() | ||
Fig. 2 Employing triazine to generate sequence-defined macromolecules on amine-functionalized sold-phase. |
Given the efficacy of N-heteroaromatic SNAr couplings this reaction has been used both in-solution and on SP for a variety of applications.30 For example, Sumerlin et al. reported the use of Trz as a BB for the post-polymerization homo- or hetero-functionalization of acrylamide backbones using amines and thiols.31 Wenschuh et al. realized on-resin peptide macrocyclization using N-heteroaromatics such as Trz.32 Trz aside, a range of halogenated N-heteroaromatics have been exploited for SP SNAr couplings including: chloropyrimidines reacting with resin-bound amines;33 resin-bound thiols coupled with 3,6-dichloropyridazine afforded aminopyrazidines after aminolysis cleavage from SP;34 and 2,4,6-triaminopyrimidines were synthesized using SP techniques.35,36
Given the advantages of both SP and SNAr methodology Grate et al. published their iterative strategy for producing SDMs with diverse side-groups using Trz-based BBs (Fig. 2).37 Prior to SP incorporation, Trz-based monomers were generated using either amines or thiols, allowing a sub-monomer's side chain to be set prior to coupling. In the first SDM coupling step, the second chloride is displaced via coupling to a resin-tethered amine. The final chloride displacement is conducted at high temperature with a diamine, affording a new primary amine for future iterations. The initial examples reported consisted mainly of hexameric oligomers with molecular weights up to 1400 g mol−1 with a set of diverse side chains (alkyl, aromatic and charged residues) which led the authors to explore side-chain interaction patterns via molecular dynamics simulations.37 All reported hexamers could be facilely purified by HPLC to provide monodisperse, sequence-defined oligomers. Recently, it has been shown that careful side-chain tailoring allows foldamer structure formation via tailored hydrogen bonding and π–π interactions, akin to secondary structure formation in Nature.38
Active or latent categorization of Trz is challenging due to the multiple, equivalent reactive sites within the ring. The first SNAr coupling to SP could be classified as “active”, however an electrophilic site still exists at the chain end. This remaining ring-bound chloride is therefore a “latent” electrophile in subsequent iterations, therefore N-heteroaromatics like Trz can be considered to be simultaneously active and latent.
Besides interest in N-heteroaromatics for their SNAr reactivity on SP and in-solution,39 there is, to the best of our knowledge, only one SDM methodology exploiting N-heteroaromatics and photochemistry. Barner-Kowollik et al. have previously exploited photo-techniques to produce SDMs,40 and this was further developed to encompass N-heteroaromatics. In 2019, the solution-phase application of photo-responsive tetrazoles for SDM synthesis was disclosed (Fig. 3).41 Using orthogonal photo-responsive ligations, a [4 + 2] Diels–Alder cycloaddition of α-methyl benzaldehydes with fumarates under 365 nm irradiation was conducted. The second coupling joins a carboxylic acid to a pyrene-pended tetrazole via activation under visible light (410 nm), forming a nitrile imine which then couples to a carboxylic acid. The authors demonstrated the formation of palindromic sequences with the highest molecular weights reported reaching 10000 g mol−1, as analysed by size-exclusion chromatography after each iterative, precise chain elongation. Aside from this example, photochemical ligations of N-heteroaromatics are a new methodology to accessing SDMs and are likely to be further developed in future.
![]() | ||
Fig. 3 Combination of N-heteroaromatic and photochemistry to generate sequence-defined macromolecules by Barner-Kowollik et al. Reproduced from ref. 41 with permission from Wiley, copyright 2019. |
Thiol–ene click chemistry techniques have been applied for sequence control,50 and maleimide-based couplings have been used for controlling sequence precision.51 Within the realm of iterative couplings, Zhang et al. reported using an iterative exponential growth (IEG) protocol, akin to epoxides (see section 2.5), to produce linear SDMs using the thiol-maleimide click (Fig. 4).52 Bis-functionalized BBs were produced with orthogonal PGs at both termini: (1) a maleimide protected as its furan DA adduct (quantitatively deprotected via retro-DA at high temperatures ca. 110 °C); and (2) an acetyl-protected thiol which is robust to high temperatures and deprotected selectively using either acidic or basic conditions. Splitting a given BB in two batches allows parallel, selective deprotections (Fig. 4). The two orthogonally deprotected batches are then merged, in a second step, and coupled via thiol-maleimide click, with the maleimide being actively consumed. Iterated splitting, orthogonal deprotection, and active coupling afforded a SDM on multi-gram scales. The use of multiple customized BBs with a considered deprotection strategy allows information to be encoded into a primary sequence, which was subsequently decoded using MS/MS techniques.52 The authors reported that IEG-based polymers could be rapidly produce e.g. at the extreme end, 7.9 g of a 128-mer with molecular weights of 27.4 kDa, as determined by size-exclusion chromatography. Zhang et al. have recently published a modified protocol allowing sequence-defined, “digital” dendrimers to be constructed using this methodology,53 expanding the scope of SDM architectures, albeit dendrimers are a specific sub-set of SDMs.54
![]() | ||
Fig. 4 Doubly protected, bis-functionalized monomers containing thiols and maleimides can be converted using iterative exponential growth methodology into sequence-defined macromolecules. Reproduced from ref. 52 with permission from Wiley, copyright 2017. |
Due to the success of this IEG approach and the robustness of the chemistry, it is possible to see alternative, iterative syntheses of maleimides and the thiol-maleimide click. No feasible limitation on sequence length, architecture or side-chain formation is foreseeable, except that provided by the given synthetic method e.g., standard limitations of a SP strategy. Alternatively, employing maleimides as dienophiles in combination with dienes like furan,55 butadiene derivatives,56 and anthracene has also produced macromolecules.57 To the best of our knowledge, DA cycloadditions to form SDMs has not been reported, except for sequence-controlled macromolecules which has recently been reviewed.51 Additionally, thermally-induced, retro-DA reactions are useful for thermoreversible materials.58
Hartmann et al. reported in 2006 using CAs for the synthesis of poly(amidoamines), via treating amine-bearing resins with succinic anhydride. Basic conditions promoted ring-opening, whilst simultaneously generating an amide bond and unmasking a carboxylic acid. The newly revealed carboxylic acid was further reacted with a diamine via active ester formation to regenerate an active amine terminus (Fig. 5).65 The reported polymers could be produced with 10 repeating units, after 20 iterations, and when coupled with a poly(ethylene glycol) chain polymers with molecular weights up to 5000 g mol−1 could be achieved after dialysis in 32% yield.
![]() | ||
Fig. 5 Iterative coupling of succinic anhydride to an amine-functionalized resin and subsequent amidation with diamines generates sequence-defined poly(amidoamines). Reproduced from ref. 65 with permission from ACS, copyright 2006. |
From this sub-monomer, two-step, iterative SP protocol many variants have been reported. Hartmann et al. reported the synthesis of sequence-defined poly(amidoamines) for reduction-sensitive, degradable DNA-delivery macromolecules.66 Alternatively, in a similar approach Wagner et al. too generated precision poly(amidoamines), but this method actively employed the CA ring-opening step prior to sequential SP addition. As such a variety of alkyl and aromatic-containing anhydrides were ring-opened with Fmoc-protected amines to afford tailor-made Fmoc-amino acid, which were sequentially added to resin via Fmoc-based peptide coupling strategies.67
Overall, the iterative aminolysis of CA has been a useful methodology to precisely grow SDMs across a range of fields. Iterative CA-diamine couplings have been used for the design of precision polymer medicinal constructs,68 for example monodisperse, cationic structures were used as non-viral polynucleotide delivery vectors.69 For further information on this topic, Wagner et al. have thoroughly reviewed the current state of the art for anti-tumour siRNA delivery.70
Thiolactones are akin to their lactone cousins, except the endocyclic O is replaced by S (cyclic esters of mercapto-acids) and 4, 5, and 6-membered rings (denoted as α, β, or γ) are most prevalent (Fig. 6).74 The chemical reactivity of thiolactones also mirrors that of lactones and ring-opening can be induced using nucleophiles, particularly amines. Thiolactone aminolysis simultaneously generates an amide bond and reveals a thiol, which may be further functionalized in either a one-pot fashion or an adjunct step (Fig. 6). Advantageously, thiolactone aminolysis allows the introduction of two separate residual groups with high atom efficiency, in some cases with click-like 100% retention of all atoms.74 In 2011, Espeel et al. reported the one-pot aminolysis and photo-induced, thiol–ene double transformation of thiolactones.75 Further development by multiple research groups expanded upon this initial report and are beyond the scope of this review, but was thoroughly catalogued by Espeel and Du Prez.4,75
![]() | ||
Fig. 6 The sequential aminolysis of thiolactones and functionalization of the revealed thiol is the foundation for a range of sequence-defined protocols. Redrawn from ref. 74 with permission from Elsevier, copyright 2014. |
With a focus on sequence-definition, thiolactone strategies to SDMs have been developed both on SP and in solution, generally using the latent reactivity of thiolactones. Thiolactone 1, which is commercially-available, has been the foundation of several methodologies given the amine can be easily functionalized e.g., via amide formation. This is best illustrated by the formation of sequence-defined oligomers on SP using 2 developed by Espeel et al.76 Firstly, 1 was modified by introduction of an acrylate side-arm to give 2, which was introduced on to an already thiol-presenting resin via thiol-Michael addition (Fig. 7). Having installed a latent thiolactone on SP, aminolysis allowed the introduction of a range of side chains through amide formation. Any synthetic strategy development is not without challenges and in this case, during the aminolysis step, it was found that liberated thiols could dimerize on the resin to form stable disulfides (Fig. 7). The authors solved the loss of reactive thiol end groups via using dimethylphenylphosphine (PhPMe2). PhPMe2 had the dual function of reducing disulfide-linked dimers in situ and catalyzing the thiol-Michael addition of 2. Overall, the authors in this initial report demonstrated tri-, tetra- and pentameric structures with molecular weights up to 1700 g mol−1 as analysed by NMR, liquid-chromatography mass-spectrometry, and high-resolution mass-spectrometry. Moreover, a range of aliphatic, aromatic and heteroatom-rich functionalities were installed in the amide side-chain. This methodology illustrates the flexibility a latent heterocycle strategy offers, especially using a single BB to rapidly generate diverse SDMs.
![]() | ||
Fig. 7 Installation of a latent thiolactone is achieved via thiol-Michael addition and subsequent ring aminolysis generates an amide side-chain and a thiol end-group allowing further iterations. Redrawn from ref. 76 with permission from Wiley, copyright 2013. |
Variants of this strategy have now been reported including: automated protocols utilising isocyanate-functionalized thiolactones (in this instance the thiol–ene reaction allowed side-chain installation);77,78 combining thiolactone chemistry in-solution with the Passerini three-component reaction (widely used in its own right to form SDMs)79–81 to convergently form SDMs on gram-scale with molecular weights up to 4 kDa,82 introduction of stereocontrol when defining side-chain identity with an automated protocol,83 and the formation of multi-functional dendrimers.84 These four examples are only selected examples of the extent of thiolactone development for the generation of SDMs with diverse architectures, and the likelihood of further advanced applications and combinations of thiolactone chemistry is high, given the rich modularity available via either active or latent modalities.
![]() | ||
Fig. 8 (A) General scheme for epoxide nucleophilic ring-opening; and (B) iterative exponential growth of epoxide-based, sequence-defined macromolecules. |
![]() | ||
Fig. 9 (A) General cyclic sulfamidate reactivity; and literature reports of cyclic sulfamidates being employed on solid-phase in (B) active and (C) latent modalities. |
CSs have been widely applied in organic synthesis especially for: heterocyclic synthesis,118 enantiopure amino acid synthesis,107 and carbohydrate synthesis.119,120 Given their properties and advantages CSs are ideal candidates to produce SDMs on SP. In particular, we were attracted to CSs as we have recently reported on the synthesis of sulfated/sulfonated oligo(amidoamines) and identified CSs as versatile building blocks to generate N-sulfated SDMs.
We envisaged iteratively applying cyclic sulfamidate heterocycles in both active and latent modes (akin to thiolactone reactivity – please see section 2.4).121 CSs had been coupled on SP in both active and latent modalities but never in an iterative fashion. For example, Lubell et al. have reported using CSs as active electrophiles to precisely install lactam rings into a growing polypeptide chain (Fig. 9B).122–125 In a typical exemplar, the methodology proceeds via direct ring opening of an ester-functionalized CS e.g. 5 on an amine-bearing resin. CS ring opening generates a secondary amine, extends the main chain's backbone and reveals a terminal N-sulfate functionality. N-Sulfate removal was affected at low pH and further heating forms a lactam ring from the condensation of the secondary amine and pendant ester (Fig. 9B). Alternatively, Cohen et al. reported glycopeptide synthesis using a carboxylate-bearing 6 as a latent electrophile. Amide coupling of 6 to an amine-functionalized resin afforded a latent CS which was attacked by thioglycosides to generate glycoconjugates (Fig. 9C).126,127
We proposed a latent strategy using novel CS 7, whereby an amide coupling would afford a latent electrophile bound to a growing macromolecular chain.121 Resultantly, a range of nucleophiles could be employed to generate ring-opening. For example, using a bis-functionalized building block, in combination with 7, realizes an iterative, step-wise AB + CD chain elongation strategy (Fig. 10A). Alternatively, we envisaged coupling an amine-bearing resin with CS BBs in an active fashion to reveal a N-sulfate serving the role of an amine protecting group. Primary amine bis-alkylation and secondary amine mono-alkylation would yield branching and linear chain growth, respectively. To allow iterative chain elongation N-sulfate cleavage would regenerate a nucleophilic amine end-group, primed for conjugation with another CS (Fig. 10B).121
![]() | ||
Fig. 10 Illustrated proposals for utilising cyclic sulfamidate building blocks in (A) a latent AB + CD and (B) an active iterative manner to generate sequence-defined macromolecules on solid-phase. |
Exploring a latent strategy, a range of amide activating agents were employed to couple 7 to an oligo(amidoamine)-presenting resin. PyBOP and Oxyma-based reagents afforded amide bond formation, but nucleophilic components were released during their activation mechanisms e.g. PyBOP expels deprotonated HOBt after initial active ester formation.121 These nucleophilic components then reacted with the latent CS generating undesired side-products (8 – PyBOP-derived and 9 – Oxyma reagent-derived) in significant quantities (Fig. 11). Despite extensive experimentation no conditions were identified which circumvented side product formation, therefore, active coupling strategies were instead investigated. To this end, amine-functionalized resins were successfully alkylated with a range of CS BBs bearing sp3 and sp2-hybridized residual groups e.g., Boc, ethyl, benzyl etc. Primary amine groups could undergo bis-alkylation to afford branched architectures, whereas secondary amines yielded linear chain growth as outlined in Fig. 10. Attempts were then undertaken to find a universal N-sulfate cleavage methodology using mildly acidic solutions and elevated temperatures. Akin to literature, cleavage was observed using pH 5.5 and 50 °C for Boc-functionalized N-sulfates. All examples with sp3-carrying residual groups could not be cleaved, despite combining pH 1 and 100 °C. This synthetic limitation prevented a wide range of CS BBs being employed when building macromolecular chains.121
![]() | ||
Fig. 11 Side product formation from ring-opening during amide coupling of 7 to amine-functionalized resin following a latent strategy. |
These challenges highlighted the issues surrounded employing heterocycles for SDM synthesis, but despite the challenges faced in realising both latent and active iterative strategies, we re-deployed the ring-opening of CSs to produce a diverse set of N-sulfated oligo(amidoamine)s using the BB actively. Given our interest in biomimicry and sulfated SDMs, the resultant N-sulfated oligomers were identified as potential glycosaminoglycan mimetics.
Boronic acids (R-B(OH)2) and esters (R-B(OR1)(OR2)) are organoboranes with two pendant oxygen-bound groups and a carbon group (sp2 or sp3 hybridised), which given an unoccupied pz orbital at boron are Lewic acids, allowing reversible formation of tetrahedral complexes (Fig. 12). This reversibility has been used for sensing,131 separation technologies,132 and cross couplings e.g. the Suzuki–Miyuara reaction.129,133 Pin- and MIDA-boronic esters are structurally diverse at boron: B(pin)'s conformation is planar trigonal whereas B(MIDA)'s is tetrahedral (Fig. 12). Most boronic ester-mediated proceed from a nucleophilic, tetra-coordinated boron species, hence B(pin) is typically more reactive than B(MIDA).134 This reactivity difference has been utilized, in a procedure akin to Fmoc-peptide synthesis, for the natural product synthesis by Burke et al. Using bifunctionalised halo-MIDA boronic esters (AB-type building blocks) a given BB can be coupled via Suzuki–Miyuara methodology to a terminal boronic acid. Pendant B(MIDA) heterocycles are deprotected to the corresponding boronic acid, allowing further iterations. Burke et al. produced pentacyclic secodaphnane alkaloids from three key BBs (Fig. 13),135–137 and have subsequently automated the process, attractive for future SDM syntheses.138
![]() | ||
Fig. 13 MIDA-boronate iterative coupling strategy and its applications to generating the secodaphnane core via iterative coupling of three tailored building blocks. Redrawn from ref. 135 with permission from AAAS, copyright 2015. |
Alternatively, Aggarwal et al. have broadly developed the lithiation–borylation iterative coupling strategy, the full details of which are beyond the scope of this review.139 Lithiation–borylation couplings proceed in three distinct steps: (1) organolithium formation from α-lithiation of a carbamate (Cb) or ester e.g. 2,4,6-triisopropyl benzoates (TIB esters) (Fig. 14); (2) electrophilic trapping of an electrophilic B(pin) BB forms a transient boronate or “ate” complex (Fig. 14A); and finally (3) the “ate” complex undergoes eliminative 1,2-metalate rearrangement to form a new C–C bond with a latent B(pin) being retained (Fig. 14A).139 One pertinent example illustrates its SDM potential as in 2014, Aggarwal et al. reported the “assembly line” synthesis of linear scaffolds containing 10 contiguous stereocentres.140 Using Li-1 alone or mixed with Li-2, chiral information was precisely installed into a growing B(pin)-capped backbone. It was shown that the methyl substitution pattern (primary structure) influenced a chain's helical propensity (secondary structure) (Fig. 14B).140 Despite its exquisite control over stereochemical transfer, lithiation-borylation requires both extremely low-temperatures and expert reagent handling to achieve the high-fidelity iterative coupling results reported. In cases where a SDM's programmed stereochemical fidelity was crucial, these synthetic limitations may also limit the achievable macromolecular weights.
![]() | ||
Fig. 14 (A) General scheme for C–C bond formation via lithiation–borylation methodology; and (B) the iterative coupling of stereochemically-rich chains. Reproduced from ref. 139 with permission from ACS, copyright 2014. |
Both of these examples demonstrate that MIDA boronic ester coupling and lithiation–borylation methodologies are ideal future candidates for producing new SDMs.
Given the breadth of heterocyclic systems discussed in this review, the development of new iterative coupling strategies utilising hetero- and carbocyclic rings could be a rich vein of development e.g., lactams, aziridines, azetidines etc. Additionally, systems exclusively applied either in solution or on SP could be transferred to the other e.g., epoxides have been well-developed in solution, therefore could be re-applied for iterative SDM assembly on SP. Finally, buoyed by the example of Barner-Kowollik et al. who applied photochemistry to generate SDMs – please see section 2.1,41 advances from the organic and organometallic chemistry communities may provide inspiration and start to percolate through into the sequence-defined polymer community e.g., organo- and metal catalysis, electrochemistry, etc.
Due to the relative youth of this field, we speculate many advances in synthesis, architecture, elemental composition, and ultimately the ways in which sequence-defined polymers will be applied.
This journal is © The Royal Society of Chemistry 2021 |