Self-replicators from dynamic molecular networks: selection, competition and subsystem
coupling
Komáromy, Dávid
IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite from
it. Please check the document version below.
Document Version
Publisher's PDF, also known as Version of record
Publication date:
2019
Link to publication in University of Groningen/UMCG research database
Citation for published version (APA):
Komáromy, D. (2019). Self-replicators from dynamic molecular networks: selection, competition and
subsystem coupling. University of Groningen.
Copyright
Other than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons).
Take-down policy
If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim.
Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons the number of authors shown on this cover page is limited to 10 maximum.
Emergence in dynamic molecular networks
Parts of this chapter were published:
D. Komáromy, P. Nowak, S. Otto Dynamic Combinatorial Libraries, in: W. Zhang, Y. Jin:
Dynamic Covalent Chemistry – Principles, Reactions and Applications, John Wiley &
Sons, Chichester, UK, 2018, pp 31-119.
10
1.1. Introduction
Complex systems are ubiquitous, from cell metabolism to the collective behavior of bird swarms, from societies to whole ecosystems. One of the striking features of these systems is emergence: most properties of the system cannot be deduced from the sum of the properties of its components.[1] This
rather unusual behavior originates from the obvious fact that the properties of a system do not arise from the elements themselves, rather from their interactions. Accordingly, the study of complex systems raises a series of fundamental questions: How do emergent properties arise? How do different levels of complexity arise? Do they appear simultaneously or rather sequentially? Is it possible to capture a tipping point where a collection of elements (molecules) begin to behave as a system? In order to answer these questions, one would rather look for a discipline which encompasses the study of several different layers of complexity.
Chemistry can be regarded as such a discipline, playing perhaps a central role among sciences.[2].
The matter of its discipline includes multiple layers of complexity between single atoms and complex biochemical machinery. Moreover, as formulated recently, the most genuine question of chemistry comprises the quest for the origin of life: how did inanimate matter transformed to life?[3] It is not hard to
see that this question is intimately connected to those concerning the evolution of complexity. Thus, if one aims at understanding complexity in general, chemistry might be one of the most adequate disciplines of choice.
The aim of this thesis is the study novel patterns of emergence (self-replication, subsystem coupling, competition, self-sorting) within complex chemical systems, more closely in the field of supramolecular and dynamic combinatorial chemistry. This introduction gives a short overview about the most important types of dynamic combinatorial chemistry (as the covalent processes enabling reaction networks) and of supramolecular self-assembly (as fundamental forms of emergent phenomena), laying special emphasis on self-replication processes. As with every instance of form-matter dichotomy, this categorization is somewhat arbitrary and certain phenomena can be treated as part of covalent as well as of non-covalent chemistries. Nevertheless, a side-by-side treatment might reveal the similarities and parallels between these two fields of study.
1.2. Reaction Networks: Dynamic Combinatorial Libraries
1.2.1. Dynamic Combinatorial Chemistry
Reaction networks are a set of chemical reactions featuring molecules which are part of more than one reaction of the set – as catalysts, inhibitors, substrates, products etc. The separate chemical processes are thus interconnected via multiple equilibria and rate laws. Clearly, even a small set of chemicals and reactions can lead to highly complex dynamic scenarios, e.g. positive and negative feedback loops, oscillations or bistability. However, such dynamics is often difficult to tackle, both
11
experimentally and theoretically. Thus, experimental model systems are needed which enable the emergence of complex scenarios but are chemically simple to control and to predict.
Scheme 1. 1. A) Examples of dynamic chemistries used in DCC: 1) Imine 2) Hydrazone 3) Acetal 4) Thio-Michael exchange. B) Properties of disulfide DCC: 1) Oxidation as irreversible and thiol-disulfide exchange as reversible step; 2) monothiols yield linear, whereas dithiols yield macrocyclic species in disulfide DCC.
In the last two decades, an approach to access such systems was developed via dynamic combinatorial chemistry (DCC), enabling the study of chemical networks called dynamic combinatorial libraries (DCLs). DCC is based on reversible covalent or non-covalent bonds and as such, can be regarded as an extension of supramolecular chemistry (based on weak and reversible interactions[4])
into the area of covalent chemistry. Dynamic bonds used in DCC fulfil a handful of requirements: a. They are reversible (low activation barrier for bond formation and breakage).
12
b. They form and break with a reasonable rate under relatively mild conditions. c. Their chemistry does not interfere with other functionalities present in the system. d. They proceed without side reactions.
e. There are conditions under which bond exchange can be stopped (“freezing”; for purposes of analysis). [5]
A handful of dynamic bonds have been developed in the past decades.[6] The most important covalent
chemistries developed are hydrazone, imine, disulfide, acetal, thioester and boronic ester exchange (Scheme 1. 1A), but numerous other chemistries have recently been in the focus of research such as Michael addition,[7] alkoxyamine exchange[8] or olefin metathesis.[9,10] Combined with the plethora of
reversible non-covalent interactions (metal-ligand exchange, hydrogen bonding, π- π interaction of electron-poor and electron-rich aromatics etc.) used in classical supramolecular chemistry, these chemistries yield a rich variety of reversible bonds based on which complex systems can be developed.
As always, there are exceptions from the requirements described before. For example, mild conditions are relative: DCCs might require large amounts of catalysts[11] or high temperatures[12]. Side
reactions in DCC are sometimes unavoidable[8,13]. Different dynamic chemistries can proceed
orthogonally to each other[14] but can also act in a cooperative[15] or an antiparallel[16] manner.
In the experiments described in the present thesis, thiol-disulfide chemistry[17] is used to generate
dynamic systems (Scheme 1. 1B). Notably, this chemistry is special as it couples an irreversible reaction (oxidation of thiols to disulfides) to a reversible one (thiol-disulfide exchange). Both processes occur between pH = 5-9 in aqueous solution at room temperature without any catalysis. Normally, generation of disulfide DCLs starts from thiols. The information content of the resulting DCLs (distribution of disulfide products) is readable in its oxidized state but the DCL can be reversibly reduced, thereby erasing the information content. From a topological point of view, monothiols give rise to linear species, whereas dithiols lead to disulfide macrocycles; more highly substituted thiols give three-dimensional cages, however, this area has been only explored scarcely.[18]
1.2.2. Dynamic Combinatorial Libraries
Dynamic Combinatorial Libraries (DCLs) are mixtures prepared from one or more molecular building blocks equipped with functional groups which are capable of reversible bond formation. Upon bond exchange and breakage, the building blocks form oligomeric molecules (DCL members). The temporal evolution of each library member is governed by kinetics of exchange and other accompanying processes, whereas their final concentrations are governed by multiple exchange equilibria. Consequently, if the concentration of a DCL member is modified, the concentration of every other member which shares a common building block with this member is changed. Therefore, a concentration
13
perturbation is transmitted through the whole DCL, thereby realizing a chemical reaction network.[19] This
feature renders DCLs responsive to external stimuli and enables systems phenomena to occur, i.e. competition for a common building block, selection of a DCL member, self-sorting etc. Notably, responsiveness is a typical systems property: it is only possible in a collection of molecules if those interact with each other.
There are multiple tools to introduce such stimuli, most of which can be classified as either chemical or physical. The former mostly rely on template effects, i.e. the introduction of a molecule which bind non-covalently one DCL member(s). The latter effects are more diverse, involving the influence of temperature, pressure, mechanical stress, electric or magnetic fields etc. As always, certain stimuli (i.e. change of solvent environment) can be regarded as both chemical and physical.
Scheme 1. 2. Schematic representation of template effects in dynamic combinatorial libraries: A) external B) internal template.
C) Schematic representation of the free energy landscape of the system upon addition of an external guest. Figures modified from references [20] and [21].
As we (including the author of this thesis) have recently published an extensive review on DCLs, including relevance from the supramolecular and materials side as well as theoretical and analytical chemical aspects, the interested reader is referred to this work for detailed information.[22] Instead, we
14 1.2.2.1. Template Effects in Dynamic Combinatorial Libraries
Template effects in DCLs can arise from interaction with an external (Scheme 1. 2A) or internal (Scheme 1. 2B) template. In the former case a DCL member forms a non-covalent complex with a molecule originally not present in the library, whereas in the latter, certain library members form non-covalent bonds with each other. Regarding the thermodynamics of the process, the interaction with the template represents a new global minimum on the free energy landscape of the system. In other words, the new equilibrium (shifted towards the product side) channels the building blocks into the formation of this DCL member at the expense of others (Scheme 1. 2C). As a result, DCL members that act as host for the template, are amplified (their relative concentration is increased). Every template effect amplification but not every amplification event is the result of a template effect; e.g. building blocks that are left behind as side products of an amplification event can form a specific new DCL member on their own. The intricate mathematical relation between template binding strength, template concentration and amplification is far from trivial and is described in great detail in a number of seminal publications.[23–29]
Figure 1. 1. A) Na+-ion templated assembly of the dynamic orthoester cryptand o-Me
2-1.1.1 from diethylene glycol (DEG) and methyl orthoacetate via acetal exchange. B) From a DCL prepared from of o-Me2-1.1.1 and tetraethylene glycol (TEG) differently sized mixed orthoester cryptands can be acquired selectively using different alkali metals as templates. Figure reproduced from reference [32].
15
1.2.2.1.1. External Template Effects
Several different types of external templates (e.g. cationic, anionic and neutral) have been described. As cationic templates, transition metals serve as a primary example.[30] More recently, however, alkali
metals have also been used in DCC to specifically access cryptand cages via acetal exchange.[31,32]
Simple diols such as diethylene (DEG) or tetraethylene (TEG) glycol were combined with orthoesters in the presence of different alkali metals (Figure 1. 1). Depending on the size match between the alkali metal and the resulting crown ether moieties, three dimensional cryptands have been obtained specifically and fully characterized by NMR, MS and X-ray crystallography. Notably, isolated cryptands (such as o-Me2-1.1.1) can engage in acetal exchange upon addition of other diols. The study showed that within the framework a suitable DCC, even alkali metals, which are mostly considered as inert from the viewpoint of metal-ligand chemistry, can exert template effects.
Anions have also been used in the same manner; remarkably, from a simple dithiol bearing a pyridine core (DTP), five macrocycles of different size (dimer to hexamer) could be amplified upon interaction with different organic polyanions (Figure 1. 2). The host-guest binding in this case relies on the ionic interaction and hydrogen bonding between the amide protons of the building block and the carboxylates of the guest.[33] Similar cases, where a DCL could lead to the amplification of a plethora of different DCL
members, have been described also for biologically relevant organic molecules serving as guests.[34,35]
Figure 1. 2. Dithiol substituted pyridine DTP can afford numerous macrocycles from dimeric DTP2 up to hexameric DTP6
macrocycles via thiol-disulfide exchange and template formation by organic polyanions of different size. Figure modified from reference [22].
16
Although most external template-directed DCLs are based on the amplification of members composed of multiple building blocks, monomolecular DCLs have also been described. In these systems, isomers of a single molecule undergoing continuous exchange constitute the DCL. Bode et al. conducted extensive studies on the dynamic exchange of bullvalene derivatives.[36] Unsubstituted bullvalene is a
fluxional molecule undergoing degenerate Cope rearrangements and as such is a dynamic mixture of more than 1 million of exchanging isomers.[37] Equipping bullvalene, with functional groups showing high
affinity towards certain guests, breaks the degeneracy and enables the amplification of a set of isomers (upon guest binding). For example, covalent linkage of two porphyrin moieties to the bullvalene core renders the molecule (BVE) suitable for complex formation with fullerenes via π-π interaction between electron-rich and electron-poor aromatic moieties (Figure 1. 3).[38] As a result, the original distribution of
the isomers can be shifted upon addition of C60, which could be detected by 1H-NMR. Additionally, a
photocleavable nitroveratryl-oxycarbonate (NVOC) group linked via an enol ether moiety enables orthogonal covalent capture (“freezing”) of the new DCL. Irradiation with UV-light results in decarboxylation, resulting in a static library of isomeric bullvalones (BVN). Similar shifting of distribution in the library compsition (and thus the NMR patterns) in boronic acid appended bullvalene DCLs were used for “barcoding” different polyols, pointing out that DCL can be also used for reading and coding chemical information.[39]
Figure 1. 3. A dynamic library of interconverting isomers of bullvalene BVE (containing a photocleavable NVOC moiety) is able
to form host-guest complexes with C60. Isomers whose structure is most suitable for guest binding are amplified. Photocleavage of the NVOC group and subsequent isomerization leads to a static library of bullvalone BVN, containing the derivatives of the best binders. Figure modified from reference [22].
17
1.2.2.1.2. Internal Template Effects: Constitutional Self-Selection
The examples described so far show that upon external template addition mainly macrocyclic species are amplified. However, the concept of an external template relies on the host-guest dichotomy – the DCC members being the former and the external templates the latter. However, one can integrate both host and guest properties into the DCL, thereby enabling host-guest complex formation among or within DCL members.
Figure 1. 4. Constitutional self-selection of C1●C2●C3 from a disulfide DCL of C1, C2, C3 and PC1, guided by sterically fine-tuned intramolecular ionic interactions. Figure modified from reference [40].
Lafuente, Alfonso et al. recently showed that in the DCL prepared from C1 (cysteine), dithiol C2 and trithiol C3, the C1●C2●C3 complex is formed with high selectivity, in spite of the huge number of possible linear, macrocyclic and caged disulfide products. The driving force of this constitutional self-selection is the formation of sterically fine-tuned complementary ionic contacts between cysteine carboxylates and
C3 amide protons, as well as cysteine ammoniums and C2 carboxylates.[41] Notably, if one of these ion
pairing sites are disabled, no selective formation of the ternary complex can be observed, as exemplified by the usage of amidated cysteine PC1 instead of C1. However, by treating a diverse DCL prepared of
PC1, C2 and C3 with C1, the ternary complex forms selectively, accompanied by the co-amplification
(vide supra) of (PC1)2 (Figure 1. 4).[40]
Self-selection enables facile (although less predictable) construction of non-trivial architectures, as the DCL members tend to maximalize the number of favorable secondary interactions, most often at the expense of geometrical strain. For example, Sanders et al. showed that dithiols bearing large aromatic cores enable access to several catenated[43] and knotted[44] structures. Dithiol NDI-2 bearing two
cysteine-appended naphthalenediimide (NDI) moieties gives rise to a 3:1 mixture of a Solomon link
(SL-1) and a trefoil knot (TK-2) in aqueous solution upon oxidation by air.[42] The tight knotting arises from
18
by the cysteine carboxylates (Figure 1. 5). The design of such systems is far from trivial and the outcome is a result of the fine interplay between charge repulsion, the hydrophobic effect and sterical factors. Such unpredictable systems can give rise to serendipitous discoveries, which are, as such, always welcome in the field of complex chemical systems research.[45]
Figure 1. 5. Dithiol NDI-2 bearing two naphthalenediimide moieties forms upon oxidation a DCL of small linear and macrocylic
disulfides, which upon intramolecular template effects gives rise to to a 3:1 mixture of Solomon-link SL-1 and trefoil knot TK-2. Figure modified from reference [42].
1.2.2.2. Physical Effects in Dynamic Combinatorial Libraries
The effect of physical factors influencing DCLs is studied to a much lesser extent than that of molecular templates. Physical factors imply diverse effects such as temperature,[46] pressure, mechanical
agitation (stirring, grinding, shaking etc.), the presence of electric[47,48] and/or magnetic fields or sound.
The impact of solvent environment (co-solvents and/or salts) is the most deeply researched area, although cations and anions from salts might also act as templates.
Physical impact on DCLs can render otherwise energetically demanding exchange pathways feasible. For example, Belenguer et al. showed that liquid-assisted grinding can induce fast exchange (bond cleavage) in disulfide DCLs even in the solid state,[49,50] whereas in solution exchange only
proceeds at a reasonable rate in the presence of thiols. Similarly, in the solution state, mechanical agitation (stirring and shaking) can open pathways toward the selective formation of self-replicating species by conferring sufficient (mechanical) activation energy to overcome the barrier of replicator formation (vide infra). Interestingly, even (ultra)sound can induce disulfide exchange, as shown recently by von Delius et al.[51]
19
Figure 1. 6. A DCL is formed from of lipophilic (red) and hydrophilic (blue) aldehydes and amines (chemical formulae not shown)
as well as hydrazide E-PyH and Ca2+ ions. Upon irradiation, Z-PyH undergoes bond idomerization and releases to Ca2+ the
solution (MeCN:H2O 3:2). Ths in turn effects phase separation due to increased ionic strength, leading to self-sorting of imines
of different polarity. Image modified from reference [52].
Light is possibly the cleanest and most selective way of inducing changes in chemical systems, including DCLs.[53–55] In a recent study, Vantomme and Lehn have shown how solvent environment, light
and dynamic bond exchange affect each other in a minimalistic DCL consisting of lipophilic and hydrophilic amines and aldehydes, as well as pyridine hydrazide PyH and Ca2+ions (Figure 1. 6).[52] Due
to its mixed polarity, the DCL is soluble in a 3:2 mixture of acetonitrile (MeCN) and water. Moreover, due to geometric restraints, the trans isomer E-PyH is capable of chelating Ca2+, yielding a complex that is
soluble in MeCN:water 3:2. However, upon UV irradiation the hydrazide moiety isomerizes to give
Z-PyH, which is not capable of metal binding due to intramolecular hydrogen bonding. Consequently, the
released Ca2+ions increase the ion strength of the solution which thus undergoes phase separation (pure
MeCN and water). As a further consequence, the imines undergo self-sorting (vide infra) due to their polarity, resulting in the disappearance of imines of mixed lipophilicity. Importantly, the process is reversible e.g. heating can restore the original product distribution. More recently, Herder and Lehn developed this concept further in the study by developing light-induced alkoxyamine exchange, showing that irradiation can induce novel, out-of equilibrium exchange reactions DCC, which are not accessible
20
via the thermal pathway.[8] These two examples show that physical effects on DCLs, although much less
studied, can introduce new connections to molecular networks, which are orthogonal to the ones accessible by pure chemical means.
1.3. Self-Assembly
Self-assembly is a general phenomenon in nature and it reaches far beyond the realm of supramolecular chemistry: it occurs on every scale from molecules to galaxies.[56] Generally,
self-assembly is defined as the “autonomous [and reversible] organization of components into patterns or
structures without human intervention.”[56] Importantly, not every formation event (e.g. the covalent
coupling of two molecules to give a new one) is self-assembly. The emphasis is rather on the transformation of a high number of smaller components into one or few highly organized structures, as well as on reversibility. Regarding chemical self-assembly, the boundaries of this discipline are far from being clear-cut. One can argue about the definition of human intervention (e.g. a collection of components with even a very high affinity toward each other would not self-assemble, were there not a chemist around who puts the self-assembling components into one flask) that of reversibility (under certain conditions, every bond formation can be rendered reversible) or about the existence of “covalent self-assembly”.[57] However, the loose boundaries of the field also mean that they are expandable, which
might enable the extension of the concept of self-assembly into remote fields of chemistry.
1.3.1. General Features of Chemical Self-Assembly
An appropriate overview of the field of chemical self-assembly reaches far beyond the scope of this thesis. Self-assembled structures form a plethora of different morphologies (Figure 1. 7), such as helices,[58] capsules,[59] fibers, vesicles[60] etc. Concerning the origin of their components, they can be
assembled from molecules of biological (lipids, nucleic acids,[61] proteins,[62] peptides[63]) or abiological
(metal-ligand complexes,[64] interlocked molecules,[65] organogelators[66]) origin. Regarding their
stoichiometry, they can be well-defined (such as certain non-covalent protein complexes[67] or DNA
origami[68]) or polymeric (such as peptide hydrogels,[69] polycatenanes[70] or mesophases[71]). The driving
forces holding together self-assembled structures can be equally diverse (hydrogen-bonding, metal-ligand interactions, π-π stacking etc.).
The first question regarding the construction of self-assembling systems is their design: Is it possible for a certain desired self-assembled architecture to design (retrosynthetically) components that will predictably result in this exact architecture? Given the diversity of supramolecular self-assembled structures, no such general design principle has been formulated.
21
Figure 1. 7. Various examples of static, thermodynamically stable self-assembled structures. A) Bipyridiyl ligand L and Pd2+
ions (M) self-assemble to give a giant discrete icosidodecahedron M30L60, as shown by single crystal X-ray diffraction analysis. Figure reproduced from reference [74]. B) Amyloidogenic peptide Ac-KLVFFAL-NH2 self-assemble into nanotubes, as shown by simulations (top) and electron microscopy (bottom). Figure reproduced from reference [75]. C) Amphiphilic small dendrimer
Ar-Ala-EO3 self-assembles into multilayer dendrimersomes, as shown by electron microscopy. Figure reproduced from
reference [71]. D) 24 molecules of engineered fusion protein FkpA-KDPGal self-assemble into a discrete porous cube as shown by modelling (top) and single crystal X-ray diffraction analysis (bottom). Figure reproduced from reference [76].
However, to certain extent the size and orientation of the components act as a “code”,[72] i. e. they
predetermine the geometry and strength of noncovalent bonds which connect them to give the final self-assembled structure. The most straightforward illustration comes from the field of metallosupramolecular self-assembly, where the orientation of ligands binding sites and metal coordination spheres might very reliably predict the shape of the final assembly.[64] If there is a certain well-defined 2- or 3-dimensional
22
closed shape with edges/planes and vertices that match ligand and metal geometry, respectively, its formation is highly probable (Scheme 1. 3). However if no such match exists, no specific assembly can form – at least not in a predictable manner (empty fields in Scheme 1. 3). It is worth noting that in the case of large assemblies small changes in the geometry or binding strength at the components level might induce a large change at the assembly level.[73]
Scheme 1. 3. Geometry of ligands and metal coordination spheres serve as a code for the geometry of the final A) 2-B)
3-dimensional metallosupramolecular assembly. Figure modified from reference [64].
From a more fundamental point of view, self-assembly processes can be cathegorized according to their potential energy surface. Regarding the components, the intermediates and the final assemblies, the latter can energetically correspond to a global minimum (equilibrium self-assembly, Scheme 1. 4A), a local minimum with a high kinetic barrier (kinetically trapped self-assembly, Scheme 1. 4B) or a local minimum with a low kinetic barrier and thus requiring constant in-flow of energy to maintain the assembled state (out-of equilibrium self-assembly, Scheme 1. 4C).[77] The former two are categorized
as static, whereas the last as dynamic self-assembly (or self-organization). Importantly, a static assembly still can be (per definitionem) reversible, i.e. it can disassemble upon certain chemical or physical stimuli. Nevertheless it resides in the assembled state if no stimuli is applied. In contrast, dynamic assemblies need a continuous supply of chemical or physical energy. Whereas chemists have studied mostly static assemblies, most biologically relevant self-assembled structures (e.g. cell membranes, microtubules etc.) are dynamic, i.e. need the presence of a fueling molecule to persist; in the absence of a chemical fuel the structure disassembles.The solvent environment plays a decisive role in the stability of self-assembled structures as solvent molecules are the most abundant ones in the system and can thus effectively compete with the assembly-forming molecules in forming non-covalent bonds. This is even more pronounced in aqueous solutions: although water forms strong non-covalent bonds with moelcules
23
of even very different polarity, it prefers to form largely ordered hydrogen-bonded networks with itself, leaving behind the hydrophobic solutes which consequently get into close proximity to each other.[78]
Hence, the so-called hydrophobic interaction is nothing but a self-sorting phenomenon (vide infra).[79]
This distinction should be taken into account in the design and characterization of self-assembled systems in water.
Scheme 1. 4. Schematic potential energy landscape A) equilibrium B) kinetically trapped C) far-from-equilibrium self-assembly.
Figure modified from reference [77].
1.3.2. Subcomponent Self-Assembly
In the previously detailed examples of self-assembly (e.g. Figure 1. 7) the components are synthesized first covalently (either via chemical or biosynthesis) and in a separate step, they engage in the self-assembly process. However, chemical synthesis and especially biosynthesis (e.g. protein expression and purification) can be lengthy and tedious; moreover, even tremendous synthetic effort does not guarantee that the obtained components self-assemble in the desired way. Therefore, a workflow in which covalent and non-covalent steps occur in the same reaction mixture are of high practical use. Nitschke and co-workers developed a method coined subcomponent self-assembly[15] in
which covalent and non-covalent bonds are formed simultaneously, in a cooperative manner. Especially, imines are highly unstable and are in equilibrium with their parent aldehyde and amine subcomponents. However, they can be stabilized by metal coordination. For example, imines of a 2-formylpyridine moiety contain a chelating motif similar to that in bipyridines and can thus chelate transition metals. Consequently, in the presence of such metals the dynamic imine equilibrium is largely shifted towards the product side. As a result, imines are stabilized and metals become chelated, i.e. covalent and noncovalent dynamic chemistries act cooperatively.
In their seminal work, Leigh et al. showed that the same concept can be used for the synthesis of macrocyclic helicates, which can be transformed into circular knots upon covalent capture. Utilizing diamine DEG-DA, Fe2+ ions, and two highly similar bis(formyl pyridyl) subcomponents BFP1 and BFP2,
24
they could obtain either a pentafoil knot[80] or a tetrameric helicate (Figure 1. 8A).[81] Importantly, the
uncomplexed imine knots were not detected during the formation of the metal-ligand complex but were only formed as part of the larger structures. Remarkably, mixing all four building blocks gives a statistical mixture of pentafoil knots, showing that dynamic exchange occurs at the level of uncomplexed intermediate imines. Mixing, however, the two separately synthesized knots, no mixed structures are detected, showing that the imines are kinetically trapped within the assembly (Figure 1. 8B).[82]
Figure 1. 8. The reaction of FeCl2 and diamine DEG-DA with bis(formyl-pyridine) BFP1 or BFP2 gives rise to the tetrameric
helicate 4H (left) or the pentafoil knot 5FK (right), respectively. B) Mixing the four subcomponents gives rise to a statistical mixture of mixed pentafoil knots, whereas a mixture of the two knotted species do not undergo subcomponent exchange. Figure modified from reference [22].
A plethora of metal-organic cages,[83] helices and grids[84,85] have been synthesized and studied using
the principles of subcomponent self-assembly. However, the broader concept has not been utilized except for the imine exchange–metal chelation pairing (with a few exceptions[86]) and thus represents a
25
1.3.3. Self-Sorting
As seen previously, if a collection of multiple suitable (sub)components (e.g. multiple metals and/or ligands) are combined, the outcome is often a statistical distribution of mixed self-assembled species. However, if the codes (vide supra) of the components are designed in a way that certain pairings become sterically or electronically highly unfavorable, whereas others are highly favorable, even a high number of (sub)components might give rise only to a few self-assembled structures among numerous possible ones. Self-sorting, defined (in a much broader sense) as the effective differentiation of self and non-self[87] occurs upon such a suitable match of intrinsically coded recognition sites.
Scheme 1. 5. Schematic representation of three main types of self-sorting. Figure modified from reference [87].
The reliability of self-sorting is also based on the intrinsic molecular codes of the components, such as size, shape, directionality and complementarity of the binding sites.[87] Furthermore, in the case of
thermodynamically driven assembly, the system tends to minimize its Gibbs free energy, i.e. the self-assembling components tend to maximize the occupancy of their binding sites. Starting from a multitude of components, self-sorting can be integrative (resulting in one specific assembly)[88] or non-integrative
(resulting to multiple assemblies). On the other hand, narcissistic and social self-sorting can be distinguished: in the former case, the same components form intermolecular bonds exclusively with each other, whereas in the latter, assemblies are constructed from different molecular components (Scheme
1. 5). Besides the field of artificial supramolecular assemblies, self-sorting phenomena are also of
fundamental importance in living systems, as it enables different self-assembling biomolecules (e.g. lipids, nucleic acids, proteins) to construct separate architectures with separate functions (membranes, helices, protein complexes).
Social self-sorting can be highly effective upon wise choice of components with pairwise orthogonal interactions. In their seminal paper, Isaacs et al. have shown that upon mixing 12 simple components
26
with markedly different binding sites and affinities (e.g. K+ and a suitable cryptand, β-cyclodextrin and
adamantane carboxylic acid etc.) only 6 out of the possible 231 different binary and ternary complexes are formed.[89] More recently, Schalley et al. studied the self-sorting behavior of organic ammonium –
crown ether complexes. Initially they found that benzo-21-crown ether-7 C7 has higher affinity towards binding benzyl-stoppered ammonium thread PhA than anthracene-stoppered AnA, whereas the opposite preference holds for dibenzo-24-crown ether-8 C8.[90] Accordingly, the four components
underwent social self-sorting to form exclusively binary complexes PhA@C7 and AnA@C8 (Figure 1.
9A). Following this initial finding, a series of bivalent crown ethers and ammonium threads were
synthesized and their self-sorting properties studied. Remarkably, upon mixing divalent 21-crown-7
C7C7, divalent 24-crown-8 C8C8 with two equivalents of heterodivalent ammonium thread AnAPhA, the
tetrameric [5]pseudorotaxane AnAPhA@C7C7@C8C8 formed exclusively (Figure 1. 9B). This selectivity is remarkable, taking into account the vast number of possible linear and cyclic oligomeric pseudorotaxanes that have not been detected among the final products. [91]
Figure 1. 9. A) Social (non-integrative) and B) Integrative self-sorting of pseudorotaxanes formed from different organic
ammonium ion axles and crown ether threads. Figures modified from reference [91].
1.3.4. Self- and Cross-Replication
1.3.4.1. Mechanistic Considerations
Replication is auto- or cross-catalytic subcomponent self-assembly. More closely, self-replicators catalyze their own formation from their subcomponents, whereas in the case of cross-replication, the
27
formation of at least one replicator is catalyzed by another. These relationships can be illustrated by DNA, perhaps the most thoroughly studied replicating molecule: First, single DNA strands are formed upon covalent linkage of their subcomponents (nucleotides). Second, single strands self-assemble into supramolecular oligomeric helices according to a very well-defined code (hydrogen bonding patterns). Finally, the complementary strand, acting as a template, lowers the barrier of strand elongation, hence it has a catalytic effect (although the process, being highly complex, needs also DNA polymerase as another catalyst to replicate at a reasonable rate). Notably, a single DNA strand is not self-replicating but it is mutually cross-catalytic with its complementary strand; however, the double strand can be considered as a self-replicator. Synthetic replicators[92] in certain sense outperform DNA, i.e. they
replicate at a reasonable rate without the need for an enzyme. Importantly, however (and this distinction holds for artificial as well as biological self-replicators), the self-replicating species is always constructed from smaller subcomponents (building blocks) and its autonomous formation is kinetically hindered. Certain self-replicating molecules are even unstable when not incorporated into an assembly.[93,94]
Regarding the type of the final self-assembled structure, self-replicators can assemble into discrete (bimolecular) or non-discrete (polymolecular) supramolecular structures. The process has certain well-defined characteristics, as exemplified for bimolecular replication (Scheme 1. 6A). Building blocks A and
B can directly react to form replicator RAB by a slow, uncatalyzed, bimolecular pathway (Pathway 1). However, the building blocks are equipped with functional groups (binding sites) which enable non-covalent binding between them. As soon as enough replicator is formed, the autocatalytic pathway becomes dominant (Pathway 2). RAB binds building blocks A and B, forming a ternary complex, thereby lowering the kinetic barrier of the covalent bond formation between A and B as a result of pre-organization. Covalent bond formation gives rise to the binary complex RAB●RAB, which dissociates into two RAB molecules, thereby doubling the number of self-replicating species. This so-called exponential replication is, however, rarely realized. Two side reactions can decrease catalytic activity: First, a self-inhibitory pathway is plausible, leading to the formation of inactive RAB*, with intramolecularly occupied binding sites (Pathway 3). Second, if RAB●RAB is a strongly bound complex, the release of catalytically active RAB might be hampered. Whereas the former pathway can be rendered unfavorable by using e.g. geometrically rigid building blocks, strong self-association of the replicator is almost inevitable (as it is a logical consequence of a stable ternary complex). Replicator efficiency can be expressed kinetically in terms of the replication order. Suppose that the reaction equation and rate law rate for replicator formation from its building blocks is
𝐴 + 𝐵 → 𝑅𝐴𝐵 and
28 𝑑[𝑅𝐴𝐵]
𝑑𝑡 = 𝑘[𝑅
𝐴𝐵]𝑟[𝐴]𝑎[𝐵]𝑏,
respectively, then the order of replication is equal to r. For exponential replication r = 1, whereas in most cases r < 1 (typically, r = 0.5), corresponding to parabolic self-replication. Two typical growth curves for exponential and parabolic self-replication are shown on Scheme 1. 6C.
Scheme 1. 6. General aspects of template-directed self-replication. A) Minimal model of a self-replicating system, showing
autocatalytic formation of replicator RAB from building blocks A and B with Pathway 1 (bimolecular), Pathway 2 (autocatalytic) and Pathway 3 (self-inhibitory). B) Minimal model of auto- and cross-catalytic pathways from a building block pool of A, B and
C, as well as replicators RAB and RCD. C) Simplified rate equations and typical growth curves for parabolic (r = 0.5, red) and exponential (r ≥ 1, blue) replication. D) Schematic growth curves of exponential self-replicating system upon addition of increasing amount of seed (pre-existing replicator) to the building block pool, resulting in gradual decrease of the lag phase and increase of the initial replication rate.
The kinetic profile of a self-replication process shows three characteristic regimes. At the beginning of the process, when Pathway 1 is dominant, replication is slow, which is expressed in a so-called lag phase. However, when Pathway 2 becomes dominant, a more rapid growth phase can be observed. Finally, when the building blocks are depleted, the replicator concentration reaches a constant value. For exponential replication, the overall growth profile is a sigmoidal curve. However, not all sigmoidal
29
curves arise from a self-replication mechanism. In order to prove that the formation of the suspected replicator is indeed autocatalytic, seeding experiments are performed: i.e. a suitable mixture of building blocks (food) is treated with a certain amount of pre-existing seed and the initial replication rate is compared with that of a non-seeded sample. Disappearance of the lag phase (bypassing Pathway 1) indicates that the mechanism is indeed autocatalytic (Scheme 1. 6D), whereas the replication order can be calculated from the initial rates of formation as a function of seed concentration.[95]
This minimal model takes only one autocatalytic system into account. However, in the presence of two templates and their corresponding building blocks, cross-catalytic pathways are also feasible. In the simplest cross-catalytic system the two replicators RAB and RAC share a common building block (A), enabling e.g. the formation of RAC from A and C, now cross-catalyzed by RAB via ternary complex
A●C●RAB (Scheme 1. 6B). It should be noted that cross-catalytic pathways do not require the feasibility of autocatalytic ones. For example, a single DNA strand cannot catalyze its own formation from its constituent nucleotides but it is mutually cross-catalytic towards the formation of its complementary strand.
1.3.4.2 Bi-and Polymolecular Synthetic Replicators
Several synthetic self-replicating systems have been developed in the last decades; the interested reader is referred to an extensive review from Kosikova and Philp.[92] The first minimalistic synthetic
self-replicator was synthesized in 1986 by von Kiedrowski. It was demonstrated that the self-complementary RNA sequence CCGCGG was able to catalyze its own formation from the activated form parent trimeric oligonucleotides CCG and CGG, utilizing EDC as coupling agent.[96] A different approach by Ghadiri and
coworkers was based on peptides forming α-helices, which self-assemble into coiled coil dimers along their hydrophobic site. The covalent chemistry utilized in this study was native chemical ligation (as this reaction proceeds spontaneously under ambient conditions and orthogonal to most functional groups present in peptides), producing coiled coil replicators (R) from electrophilic thioacid (E) and nucleophilic thiol (N) fragments.[97]
In a subsequent study,[98] mutations were introduced into the nucleophilic fragments to produce two
replicators R1 and R2. From the corresponding nucleophilic peptides N1 and N2. Intriguingly, besides showing autocatalytic formation, both replicators were able to cross-catalyze each other from their corresponding building block (Figure 1. 10A). This scenario is the first example of symbiosis in artificial self-replicating systems and as such, paved the way towards the study of replicator ecologies in artificial systems. Ashkenasy et al. conducted numerous studies on replicators based on the coiled coil design, studying systems highly relevant from the perspective of artificial living systems, e.g. the spontaneous emergence of mutations,[100] light-triggered replication,[101] or bistability.[102]
30
Besides peptides and nucleic acids, artificial self-replicators have been synthesized based on fully abiotic binding motifs, although these were also inspired by multivalent hydrogen bonding present in nucleic acids. The group of Rebek pioneered the study of replicators of non-biological origin, based on Kemp’s acid binding motifs.[103–105] Although self-replication was effective, the synthesis of the monomers
required much synthetic effort.
Figure 1. 10. A) Minimalistic self-replicating system based on abiotic molecules: Coiled coil peptides R1 and R2 are formed
from their respective building blocks E, N1 and N2 (sequences shown on top) via native chemical ligation (left), through auto- and cross-catalytic pathways (right).[98] B) Maleimide CM and nitrone APN form oxazolidine OXZ via [3+2] cycloaddition, which
can effectively bind their constituent building blocks via hydrogen bonding. Due to preorganization, the kinetic barrier of the coupling reaction is lowered, leading to effective autocatalytic formation of OXZ.[99]
Philp et al. developed a highly selective, modular and fully abiotic minimalistic self-replicating system,[99] based on earlier work from von Kiedrowski and Sutherland.[106,107] The approach was based
on hydrogen bonding as the non-covalent binding motif and alkene-nitrone [3+2] cycloaddition as the covalent reaction leading to the replicators. Aminopyridine-conjugated nitrone APN and carboxylic acid-appended maleimide are capable of forming an oxazolidine OXZ via a slow, uncatalyzed pathway. However, the two building blocks can bind to each other as well as to OXZ via complementary hydrogen bonds, self-assembling into the ternary complex OXZ●APN●CM. Preorganization of the building blocks accelerates the cycloaddition, i.e. OXZ is able to catalyze its own formation from its building blocks (Figure 1. 10B). This methodology was extended towards the construction of self-replicating rotaxanes[108], and systems showing symmetric[109] and asymmetric[110] cross-replicating behavior.
Moreover, by coupling the system to a simple nitrone DCL, replicators were able to direct the selective synthesis of their own building blocks, thereby depleting the library from the needed building blocks.[111]
31
During the past decades, effective (exponential) self-replication turned out to be highly challenging, due to the ineffective dissociation of the replicator-replicator complex. One way to overcome this problem was the introduction of polymeric self-replicators, which self-assemble into supramolecular polymers, and thus always possess non-complexed (occupied) replicator sites. Our group has conducted extensive studies on polymolecular self-replicators, based on disulfide DCLs as exchange pool and association of peptides into β-sheet structures as non-covalent interaction (Figure 1. 11). The building block LK is an aromatic dithiol appended with a pentapeptide side chain with alternating hydrophobic and hydrophilic amino acids (typical structural motifs of β-sheet peptides). Upon stirring in aqueous solution in the presence of atmospheric oxygen, initially, small disulfide macrocycles (trimers and tetramers) are formed as main products via oxidation and building block exchange, whereas larger macrocycles are initially present in minor amounts.
Figure 1. 11. Peptide building block LK gives rise to hexamer macrocycles upon exponential self-replication enabled by a fiber
elongation-breakage mechanism.[94,95]
Each macrocycle has a certain tendency to form non-covalent aggregates with each other. The Gibbs free energy contribution to these interactions originates from hydrogen bonding (between peptide chains) as well as from hydrophobic effects. Consequently, the more of these binding motifs are present in an oligomer, the higher is its tendency toward self-assembly into a well-defined supramolecular structure. Smaller macrocycles, although abundant, are not capable of forming sufficiently strong non-covalent bonds. In contrast, larger macrocycles aggregate much stronger but they are not abundant enough. In an ideal case, there is a tradeoff between these two factors; more specifically, for the LK building block, the macrocyclic hexamer is able to self-assemble into small well-defined stacks (primary nuclei). This process has a high kinetic barrier (uncatalyzed pathway), i.e. the amount of hexamer stacks increases
32
slowly. These nuclei, however, grow autocatalytically, i.e. promote hexamer formation by binding to monomers and small macrocycles, thereby forming longer stacks. After reaching a certain size, these stacks become susceptible to breakage upon the impact of mechanical energy, which in this particular case is conferred by shear forces arising from stirring. Upon breakage, the number of catalytic fiber ends is redoubled, resulting in the exponential increase of catalytic sites after every breakage cycle. These new, shorter fibers (secondary nuclei) are capable of growth until the total amount building blocks is depleted.[94,95,112]
This mechanism has several remarkable aspects. First, the hexamer is not very stable in the unassembled state, i.e. it is not the thermodynamic but the kinetic product of the system. However, in this case, non-covalent self-assembly can direct covalent bond formation toward the exclusive formation of an otherwise unfavored molecule. (Scheme 1. 4B). Second, as the fibers are linear (not toroidal), they always have two free fiber ends, i.e. the process is not hampered by self-inhibition nor strong replicator-replicator binding. Indeed, upon mechanical agitation, replicator-replicator growth is exponential, as proven by detailed mechanistical studies.[95] Third, less hydrophobic peptide sides contribute less binding energy,
thereby requiring in larger ring size of the replicating species to hold the resulting assembly together.[112]
Solvents, on the other hand can enhance hydrophobic interactions and accordingly dictate replicator size.[113] Moreover, the size distribution of the fibers can be tuned by higher, controlled shear stress,
giving rise to supramolecular polymers with uniform distribution and thus well-defined properties.[114]
Finally, mixed DCLs or cross-seeding experiment were observed to give rise to highly complex behavior, such as replicator diversification[115], parasitism[116] or history dependent emergence of replicators[117],
reminiscent of typical ecological scenarios and thus highly relevant to origin-of-life and de novo life research.[118,119]
Notably, only a small number of conceptually different polymeric self-replicators have been reported.
[10,93,120] These self-assemble into micellar structures and detailed mechanistical information about their
formation (e.g. replication order) is scarce. Therefore, the peptide replicator system described above offers a well-suited platform towards further development, such as replicators with catalytic function or increased information content.
1.3.5. Self-Organization
The examples above illustrating the multiple facets of self-assembly phenomena have one common trait: they occur under no or only temporary inflow of energy. However, most processes occurring in living organisms require different conditions. For example, in living organisms, many self-assembled structures emerge only as long as a constant supply of chemical or physical energy is conferred to the system; upon interruption of this supply the system undergoes disassembly.[121] In the field of systems chemistry
33
self-organization (or dynamic self-assembly). Upon assembly, energy is dissipated; therefore, the term
fueling has been commonly used in this context. More specifically, upon physical fueling (e.g. with light
or mechanical agitation) heat is produced, whereas upon chemical fueling the fueling molecule induces assembly by a series of reactions and is thereby turned into a by-product (waste). Importantly, the assembly and disassembly pathways are not the exact opposites of each other; in other words, upon disassembly, waste is not transformed back into the chemical fuel at a reasonable rate (Scheme 1. 4C). Despite its generality, chemical self-organization phenomena have been studied in depth only in the recent years. The underlying reason is twofold: First, in contrast to equilibrium thermodynamics, the boundary conditions for such systems (i.e. which self-organized structure and pathway is selected from among multiple possibilities) are still debated.[122] Second, chemical fueling needs at least two different
reactions occurring simultaneously (or in short succession), one leading to covalent coupling, the other resulting in covalent bond breakage; such pairs of reactions are scarce.
Figure 1. 12. A) Structure of the dominant form of building blocks DC-OH (charged) and DC-OMe (neutral) in aqueous solution
at pH = 10.7. B) Schematic representation of the dimethyl-sulfate (DMS)-fueled self-organization: upon treatment of DC-OH with DMS, charge neutral DC-OMe is formed and methyl sulfate anions are released (left). DC-OMe self-assemble into nanoscale fibers (middle). The ester bond is slowly hydrolyzed in aqueous solution, releasing methanol and DC-OH, whereby the fibers disassemble (right). C) The fibers assembled from DC-OMe form a fibrous network, resulting in fast (1 hour after the addition of DMS) gelation of the sample. On a longer time scale (12 hours) the gel is destroyed as the fibers disassemble. D) Fluorescence microscopy images highlighting growth (0.3-1 hours after DMS addition) and disassembly (5.3-7.1 hours after
DMS addition) of the fibers. Figure modified from reference [123].
In recent years, several systems have been studied which demonstrated the application of light,[124,125] mechanical agitation,[122] electric current[126,127] as well as small biological[128] or abiological
34
molecules[129] to fuel the temporary assembly of nanoparticles,[130] gels,[131] interlocked molecules,[132]
nanocapsules[133] or fibrous assemblies.[134] A recent, rather simplistic system made use of the fact that
ester bonds can be formed quickly under aqueous conditions but they disassemble in the presence of a base.[123] The authors used a simple dipeptide conjugate (N-dibenzoyl-cystine, DC-OH) as a precursor.
Under basic conditions (pH = 10.7), DC-OH is present in a molecularly dissolved form, as charge repulsion between carboxylate moieties hampers any possible self-assembly (Figure 1. 12A). However, upon treatment with excess dimethyl sulfate (DMS) the charges get neutralized upon formation of the methyl ester DC-OMe. Owing to its relatively large aromatic surface, DC-OMe undergoes rapid self-assembly into fibers of several hundred nanometer in length. Remarkably, the fibers self-assemble into a fibrous network, resulting in gelation of the sample (Figure 1. 12B,C). However, the ester bonds are unstable at this pH and thus DC-OMe is slowly hydrolyzed back to DC-OH and methanol (waste), resulting in disassembly of the gel (Figure 1. 12C,D). Several gelation-disassembly cycles could be performed; moreover, under constant inflow of the fuel, the self-assembled state can be maintained. Upon tuning of the conditions (concentration of the reactants as well as pH), the lifetime of the gel could be tuned. Importantly, as indicated above, DMS is not produced in the disassembly process; rather, the net reaction is the hydrolysis of DMS to sulfate ions and methanol, catalyzed by DC-OMe.
This system thus illuminates several rules to be used in the design of self-organizing chemical systems: First, the chemical reaction triggering assembly has to be faster than disassembly. Second, a assembling catalyst is a good candidate for a organizing material, given that it is capable of self-assembly in the active state but not in the dormant state.[135] Despite, however, intriguing examples from
the recent years, such design rules are far from being general enough to construct robust and tunable dynamic self-assembling systems.
1.4. Subsystem Coupling
The previous examples from the field of self-assembly demonstrated how careful design of molecular codes can give rise to complex supramolecular structures. However, the demonstrated systems were studied on their own, i.e. without interaction with other systems. In contrast, in most complex systems observed in nature, self-assembled systems rarely operate in isolation. Rather, the various subsystems, featuring a specific (functional) assembly on their own, are connected by messenger molecules, which can exit from one subsystem, enter another one (specifically) and modulate the assembly mode or function of the latter. In nature, series of coupled subsystems are wired up to give signaling cascades which can transport and transform small molecular signals to relatively high distances.[136]
We have already detailed in the context of DCC how small molecules could be similarly “wired up”: as the concentrations of DCL members are determined by multiple equilibria, the perturbation in the concentration of a specific DCL member (e.g. by guest binding) most likely affects that of the other
35
members as well.[27] However, affecting the distribution of DCL members by shifting one equilibrium
cannot be regarded as subsystem coupling as the boundaries of separate subsystems are not well defined. We define here a subsystem as a collection of molecules which adopt a well-defined supramolecular assembly form or possess some characteristic function (e.g. catalysis, guest binding).
Coupled artificial subsystems are ubiquitous in molecular biology. In the context of prebiotic chemistry, the coupling of simple information-rich molecules and compartments have been studied in the recent years.[119] However, the exploration of coupled sytnhetic subsystems has been restricted to a
handful of examples.[137–139] The challenge is obvious: the subsystems have to maintain their own
integrity, i.e. self-sort and not exchange components, whereas at the same time they should provide signaling molecules which are able to interact with other (but not all) subsystems.
Figure 1. 13. A) Overview of a signaling cascade enabled by subsystem coupling: addition of C18-substituted norbornadiene
18N to tetrazine-containing cube Fe8(TB)12 produces pyridazine-edged cage Fe8(PB)12 and releases cyclopentadiene 18N, which undergoes a second reaction with polar maleimide-edged tetrahedron Fe4(PM)6 to give apolar Fe4(18P)6. B) Fe4(18P)6 is soluble in cyclopentane and is able to transfer its polar cargo PF6- into the apolar phase. Figure modified from reference [140].
Nitschke et al. have recently devised, starting from fully abiological molecules, a system in which subsystems are coupled via a messenger molecule akin to those in regulatory networks in cells (Figure
1. 13).[140] A stoichiometric mixture of tetrazine-based dialdehyde TA, para-tert-butylaniline BA and
Fe2+ions give rise to the cubic coordination cage Fe
8(TB)12 via subcomponent self-assembly. Similarly, bipyridyl-based dialdehyde PA and maleimide-appended aniline MA undergo subcomponent self-assembly with Fe2+ions to give tetrahedral cage Fe
4(PM)6. Notably, despite the similarity of the two aldehyde building blocks, a mixture of the two cages does not allow for the formation of mixed cages due to the different geometry of the aldehyde groups (geometric codes). Furthermore, both cages are
36
capable of binding anions (e.g. AsF6- or PF6-) in moderately polar solvents, e.g. acetonitrile/chloroform
mixtures. The addition of norbornadiene 18N substituted with C18 alkyl chain triggers a cascade of reactions. First, 18N and tetrazine-based edges of Fe8(TB)12 undergo an inverse electron-demand Diels-Alder reaction to give cage Fe8(PB)12 with pyridazine edges. Second, in this reaction a stoichiometric amount of C18-substituted cyclopentadiene 18C is released. Third, this by-product reacts in a subsequent cycloaddition with the maleimide end-groups of ligand PM in cage Fe4(PM)6. As a result, the originally polar Fe4(PM)6 is transformed to apolar Fe4(18P)6 (Figure 1. 13A). The newly formed cage is soluble in cyclopentane (immiscible with the original solvent mixture) and thus could transfer the encapsulated cargo PF6- (otherwise only soluble in polar solvents) into the apolar phase (Figure 1. 13B).
This example demonstrates that upon careful design of the subsystems molecular signals can be amplified to induce new emergent phenomena at systems level (in this case macroscopic transport), paving the way for synthetic signal transduction systems.
Figure 1. 14. A) Representation of a feedback loop based on coupling three subsystems, i.e. a positive feedback loop (enzyme
E1* catalyzing its own formation from precursor E1) with a negative feedback loop (E1* catalyzing the formation of its own inhibitor
Inh from ProInh), and a delay module (E1 catalyzes the degradation of ProInh to an intermediate inhibitor, which is transformed
to Inh by catalysis with another enzyme E2). B) The system under study featuring trypsinogen (Tg), trypsin (Tr), aminopeptidase (Ap), KQ-ABF or KQ(Me)-ABF and ABF as E1, E1*, E2, ProInh and Inh, respectively. C) Kinetic profiles of batch reactions of the systems under study, showing too early inhibition of and thus constant decrease of [Tr] for pro-inhibitor KQ-ABF (red traces), as well as the expected spike in [Tr] for KQ(Me)-ABF (black traces). D) Schematic view of the flow reactor used in the study, showing separately regulated inflows of reagents, as well as outflow of the reaction mixture. E) Tr shows an oscillatory concentration profile in time under optimized flow rates. Figure modified from reference [143].
37
The above example shows that the coupling of subsystems can occur in a unidirectional fashion, i.e. subsystems affected later in the cascade process do not exert any influence on those affected previously. However, if subsystems can modulate each other mutually, feedback loops can emerge, which serve as a foundation for regulatory elements in living systems.[141] Indeed, in cell biology a whole taxonomy of
linear and circular network motifs can be constructed from appropriate coupling of positive and negative feedback loops.[142] However, such networks have been scarcely described in the context of synthetic
chemical networks.
One recent example is reported by Huck and coworkers and it describes the construction of a simple system showing oscillatory behavior, enabled by the appropriate coupling of subsystems in the form of positive and negative feedback loops (Figure 1. 14A).[143,144] The system (Figure 1. 14B) used in the
study is based on the observation that trypsin (Tr), itself a polypeptide, can catalyze the hydrolysis of peptide bonds. The first subsystem consists of a positive feedback loop, in which trypsin catalyzes the hydrolysis of the protein trypsinogen (Tg) to produce more trypsin. The second subsystem is based on a negative feedback loop, arising from the ability of trypsin to catalyze the formation of its own inhibitor 4-[2-aminoethyl] benzenesulfonyl fluoride ABF, by hydrolyzing peptide-conjugated precursor KQ-ABF. Combining Tg and KQ-ABF with a catalytic amount of Tr would lead first to the autocatalytic production of Tr (increase of [Tr]). This would result in increased rate of hydrolysis of KQ-ABF, producing ABF, which in turn sequesters the Tr as Tr●ABF complex (decrease of [Tr]). However, the expected kinetic profile was not observed as the rate of inhibition was too high (Figure 1. 14C, red traces). Therefore, a modification of the inhibitor was implemented by introducing a third subsystem, which resulted in the delay of inhibitor production. Specifically, methylated K(Me)Q-ABF was used as pro-inhibitor, which is first cleaved by Tr, and the resulting Q-ABF was cleaved by a second enzyme, aminopeptidase (Ap). The rate of both new reactions can be tuned independently from each other, resulting in sufficient delay time to observe a spike (increase, then decrease) in [Tr] under batch conditions (Figure 1. 14C, black traces). Moreover, under suitable flow conditions (separately tuned inflow rates of Tg, Tr, Ap and
K(Me)Q-ABF, as well as outflow of reaction mixture, see Figure 1. 14D, red traces), the concentration
of Tr was observed to oscillate in time (Figure 1. 14E). Remarkably, the authors measured the rate laws for each separate reaction step and subsequently used simulations to find a suitable part of the parameter space, in which the expected systems behavior would occur. Thus, the study demonstrated that the design of subsystem coupling can be performed in a modular manner, can be controlled quantitatively and by exploring the parameter space, can give rise to different temporal patterns. Indeed, the same methodology has been used more recently to construct a similar oscillating chemical system, however, at this time constructed from small, mostly abiotic organic molecules, proving the generality of the concept outlined above.[139]
38
1.5. Content of this Thesis
It would do unjustice to the nature of scientific discovery if we tried to introduce the chapters of this thesis as being carried out according to a previously defined scientific plan. Rather, they were conceived as partially serendipity-, partially association-based consecutive pieces of research and are also introduced in the following as such.
Furthermore, in this thesis we do not aim at mimicking living systems and their Darwinian evolution as the foremost and exclusive goal of our investigations. Rather, we regard living entities as the most advanced examples of complex systems – nevertheless, just examples. With this notice we could free ourselves from the ideological burden of having artificial life and Darwinian evolution as the sole focus of our investigations and instead, we can focus on the intriguing molecular and supramolecular phenomena that our chemical systems are able to produce.
In Chapter 2 we describe a dithiol building block which upon oxidation and oligomerization can engage in two different mode of self-assembly, depending on physico-chemical factors. In the absence of mechanical agitation, a diverse collection of unprecedentedly large macrocycles is formed which self-assemble into microdroplets. In contrast, agitation induces the formation of a hexameric self-replicator, assembling into nanoribbons. Our results suggest that in this specific case non-covalent self-assembly directs covalent bond formation in two different manners, towards molecular diversity in one case and molecular specificity in the other.
In Chapter 3, the self-replicator synthesized and studied in the previous chapter is taken as a subsystem and connected via disulfide DCC to another subsystem, capable of effector recognition. Specifically, the novel dithiol building block is mixed with a previously studied dithiol, which can oligomerize to form i.a. a strong spermine receptor in the form of its tetramer. Mixing the two monomers in a stirred solution results in a DCL with no preferred species; however, spermine (effector) addition yields the tetrameric receptor in high selectivity and consequently, leaves behind an almost pure DCL constructed from the first building block. This DCL, upon further agitation, produces selectively the hexameric replicator. The process is modular and its quantitative parameters can be tuned by the addition of effector added.
In Chapter 4, the replicator explored in Chapter 1 and the previously studied peptide-based self-replicators are compared. Starting from mixed building blocks, mixed self-replicators emerge, according to the statistical distribution corresponding to their stoichiometry. Whereas no self-sorting of replicators take place at the molecular level, at the supramolecular level transition from one morphology to the other is not continuous, i.e. incorporating even low amounts of one building block in the replicator assemblies of the other disrupts the original supramolecular organization, resulting in a third type of nanoscale morphology, more resembling to that of the peptides. Accordingly, we observed that the peptidic and the mixed self-replicators are mutually cross-catalytic, whereas the novel self-replicator described in Chapter
