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.
Chapter 5
Stoichiometry
dictates
self-assembly
pathways: from fiber bundles to daisy chains
This chapter will be submitted for publication:
D. Komáromy, T. Tiemersma-Wegman, J. Kemmink, J. Ottelé, F. S. Aalbers, G. Portale,
I. Marić, G. Monréal Santiago, P. W. J. M. Frederix, V. Saggiomo, P. van der Meulen, B.
Liu, S. Otto manuscript in preparation
162
5.1. Introduction
Biomolecules are synthesized in living systems from a limited number of different building blocks in
a stoichiometric, sequence and stereospecific manner, resulting in a vast number of biooligomers and
-polymers. These molecules do not only exhibit precisely tunable molecular complexity (and the ability
to organize into astonishingly complex supramolecular architectures in a hierarchical manner), but also
fulfil myriads of biochemically and biologically relevant functions. For example, nucleic acids are able to
replicate themselves sequence-specifically (including error-correction mechanisms to counterbalance
mutations) to give polymers with a molecular weight of several megadaltons from only four small building
blocks and transmit hereditary information confidently on a time scale of several million years. As another
example, enzymes are assembled in a sequence- and stereospecific manner from only a handful of
amino acid building blocks to afford subtly tuned biocatalysts (composed of several hundreds of amino
acids) for complex chemical transformations, whose specificity is not only determined by the correct
assembly of the active site but also by the rest of the protein, responsible for conformational flexibility
and supramolecular organization. The latter point is especially important as the ability of proteins of
several hundreds of kilodaltons to self-assemble into higher ordered structures can be triggered by a
mutation of a single amino acid.
[1]Uncovering the origin of these stoichiometry- sequence-and stereospecific assembly processes is a
highly relevant and still underexplored research area within the field of origin--of-life research.
[2]On the
one hand, non-selective oligomerization of nucleotides and amino acids under prebiotically plausible
conditions have been reported.
[3–7]On the other hand, sequence-selective copying of oligonucleotides
has been demonstrated, however, only on a pre-existing oligonucleotide (primer) in the presence of a
template.
[8–11]The so-called minimotif hypothesis
[12]is one of the few theories according to which
oligomerization and sequence-selection could have proceeded simultaneously. According to this theory,
small peptides, which could be formed from pre-existing amino acids
[13](i.e. under alternating wet-dry
conditions
[14]) are already capable of forming relatively strong non-covalent complexes with each other.
Moreover, even very short (but sequence-specific) peptides are able to catalyze simple reactions. These
two phenomena might serve as a basis for selection for small peptides given the continuous formation
and hydrolysis of amide bonds in the primordial soup, representing a primitive peptide evolution scenario
(involving replication and selection). However, determining the exact conditions under which the
aforementioned assembly pathways were developed, is far from being trivial. Thus, instead of attempting
to gain insight into the exact pre-biotic scenarios of the past, another approach has been developed in
the recent years, involving the creation of life-like systems from fully or partially abiotic molecules. In the
recent years, so-called de novo life research has given better insights regarding mechanisms of
self-replication
[15–17]catalytic properties of biologically relevant supramolecular structures
[18–23], the formation
163
of compartments
[24], the dynamics of dissipative processes
[25,26], oscillations
[27]and transient
assemblies
[28–30]processes, all being essential features of living systems. However, capturing or
mimicking the first event of stoichiometry- and/or sequence-selective self-assembly of biologically
relevant molecules has so far remained elusive.
In order to simulate the prebiotic situation in which numerous different building blocks had the chance
to form covalent bonds (which can be broken later) with each other, we and others developed dynamic
combinatorial chemistry (DCC) as a platform for such processes. In DCC, reversible bonds
[31,32](e.g.
hydrazone, disulfide, imine, metal-ligand) are formed and broken simultaneously, thus giving rise to
highly complex equilibrium mixtures (dynamic combinatorial libraries, DCLs) from a small number of
building blocks. Non-covalent interactions between an external guest and one of the DCL members or
among DCL members can shift this equilibrium toward the formation of a specific oligomeric DCL
member. In the former case, the process results in a discrete host-guest complex, whereas in the latter
case, self-templation leads to supramolecular polymers, which can nevertheless give rise to various
well-defined nanoscale aggregates (i.e. fibers, nanoribbons, nanotubes, vesicles).
[33]Such supramolecular
architectures can be further classified, based on the activation barrier of their formation, which can be
relatively low (compared to that of the formation of other DCL members) or relatively high. In the latter
case, the formation kinetics of the preferred oligomer is initially slow due to the high activation energy
(lag phase), but after reaching a critical concentration, the self-templation process enables its fast
formation in an auto-catalytic manner. Thus, overall, the system exhibits self-replication as an emergent
property.
[34]In the recent years, our group explored the self-replication properties of disulfide macrocycles based
on oligopeptide-conjugated dithiol building blocks,
[35–38]such as 1 (Scheme 5. 1). Upon oxidation by
atmospheric oxygen in a stirred aqueous solution, building block 1 forms an equilibrium mixture of
disulfide macrocycles. The octameric macrocycle 1
8, which is a thermodynamically unfavored product
(due to its high entropy of formation) is, however, capable of templating its own formation from the
exchange pool, thereby forming stacks of macrocycles.
[36,37]The single octamers are held together by
hydrogen bonding and London dispersion forces between the alternating hydrophilic and hydrophobic
amino acid units, respectively, leading to a beta-sheet like arrangement of the peptide chains. The initial
growth (nucleation) of the stacks (fibers) is slow, resulting in a lag phase in octamer growth kinetics.
However, upon reaching a certain length, the fibers become susceptible to breakage caused by
mechanical agitation (stirring), resulting in the redoubling of fiber ends, which are the sites of growth in
the assembly. This leads to an exponential growth of the octamers in a continuing breakage-growth
cycle, finally depleting the DCL material by converting it into the self-replicating species.
[39]164
Scheme 5. 1. Dynamic Combinatorial Library Made From Building Blocks 1 and 4 Can Give Rise to 18 (via Self-Replication),
1442 (via Self-Assembly)or (1143) 3 (via Threading).
Recently we have reported
[40,41]that in a 2-component disulfide DCL, self-replication can be triggered
by an effector molecule, i.e. by an external template, which can shift the equilibrium by sequestering the
oligomers constructed from one building block, leaving behind solely oligomers of the other building
block, one of which is capable of self-replication. This system thus links the formation of discrete
host-guest adducts with that of self-replicators in a consecutive manner. However, the simultaneous formation
of distinct architectures, based on different principles of self-assembly (stoichiometric host-guest
chemistry and self-replication) in the same system have still not been realized in the context of DCC.
Presumably, this scenario is more likely if the system is given the capability of producing more,
165
characteristically different oligomers, each of which nevertheless being capable of self-assembly. Such
a situation would thus enable the formation of distinct species originating from the same pool of building
blocks, highly reminiscent of the aforementioned, sequence-specific formation of biooligomers from their
constituents.
Moreover, Vial et al. recently reported
[42]that the tetrameric disulfide macrocycle formed from building
block 4 (2,5-dimercaptoterephthalic acid) forms host-guest complexes with primary amines, including
lysine-containing small peptides. These host-guest complexes are based on the strong hydrogen
bonding and ionic interactions between the protonated primary amino group of the lysine unit and the
carboxylate groups of 4 (Scheme 5. 1, bottom right). In line with these observations, our group had
previously reported
[43]that the formation of the same tetramer (4
4
) can be templated by the addition of
spermine (Sp) and various α,ω-diamines via the formation of a highly stable host-guest complex; here,
additional binding energy is provided by the hydrophobic effect, whereby the alkyl moiety of the
corresponding diamine is buried by the aromatic core of 4
4. Thus, we envisaged that a DCL composed
of building blocks 1 and 4 would represent an interesting scenario where a molecular fragment of one
building block (i.e. the lysine side chain of 1) serves as a guest for oligomers composed of the other
building block (4). As a consequence, we expected the formation of mixed oligomers featuring both donor
(lysine) and acceptor (aromatic carboxylate) units, presumably possessing different capabilities for
self-assembly.
Here we report that in line with the expectations outlined above, a dynamic combinatorial library
composed of 1 and 4, from which (based on the initial building block ratio) three different species with
markedly different self-assembly properties can emerge: the octamer 1
8which self-replicates by forming
fibers; the mixed hexamer 1
44
2which also self- by forming fiber bundles; or the mixed tetramer 1
14
3which self-assembles giving a discrete molecular product, i.e. a [c3] daisy chain pseudorotaxane
(Scheme 5. 1).
5.2. Results and Discussion
5.2.1. Selective formation of 3 different molecules in a 2-component DCL
In order to explore possible different self-assembling species arising from a DCL composed of
building blocks 1 and 4, we prepared 24 libraries in aqueous phosphate buffer (pH = 8.15, 50 mM) each
featuring the same total amount ([1] + [4] = 2.0 mM) but different molar ratio of the two building blocks,
ranging from 0 to 100 mol% of 4. The libraries were stirred at 1200 rpm at room temperature and
monitored with UPLC, until no further change was detected in the DCL composition. In libraries
containing no or only low amounts of 4 the octamer 1
8emerged (Figure 5. 1A), as expected. Increasing
166
the amount of 4 to 10 mol % led to the disappearance of the octamer, without the emergence of any
preferred species (Figure 5. 1B). Surprisingly, further increasing the relative concentration of 4 to 40
mol% led to the appearance of the mixed hexamer 1
44
2(Figure 5. 1C). At even higher relative amounts
of 4, the mixed hexamer persisted, however, the mixed tetramer 1
14
3appeared (Figure 5. 1D). The
oligomers 1
44
2and 1
14
3coexist up to 75 mol % of 4, at which point the mixed hexamer is not present
any more. At even higher relative concentrations of 4 the mixed tetramer emerges as a preferred species
(Figure 5. 1D), whereby in DCLs with > 80 mol% 4, the relative amount of the mixed tetramer gradually
decreased, giving rise to 4-only oligomers. (See details on Figure S5. 58).
Figure 5. 1. UPLC chromatograms of DCLs prepared from 1 and 4 ([1] + [4] = 2.0 mM) in aqueous phosphate buffer (50 mM,
pH = 8.15; for A)-E)) or aqueous acetate buffer (50 mM, pH = 3.2; for F) ) containing A) 0, B) 10 C) 40 D) 60, E) 80, F) 33 mol % 4.
Plotting the relative amounts of the three distinct self-assembling species against the mole
percentage of 4 (Figure 5. 2) reveals that at molar ratios close to the stoichiometry of each preferred
species, the corresponding molecule constitutes ca. 90 mol % of the DCL (0, 35 and 80 mol % of 4 for
167
1
8, 1
44
2and 1
14
3, respectively). This trend indicates that under the specified stoichiometries, each of the
three macrocycle is the most stable species in the system (either kinetically or thermodynamically).
Figure 5. 2. Relative amount of preferred species 18, 1442 and 1143 (in DCLs prepared with increasing amounts of 4 ([1] + [4] =
2.0 mM) in aqueous phosphate buffer (50 mM, pH = 8.15).
5.2.2. Structure and assembly of the mixed hexamer 1
44
2Octamer 1
8is a well-known self-replicating species, described by our group previously.
[36,37]Therefore, we turned our attention first towards the mixed hexamer 1
44
2. We proceeded in the
characterization of this oligomer in a bottom-up manner, first by determining its molecular structure,
followed by its supramolecular organization and finally, by studying its kinetics of formation. Initially, it
should be noted that under the conditions which prefers the formation of this hexamer, no other hexamers
(e.g. 1
54
1or 1
34
3) were formed in substantial quantities (see Scheme 5. 1 and SI Section 5.4.5). In other
words, in this DCL not only the macrocycle size but also the stoichiometry is selected for. In order to get
better insight into the origin of this specific assembly, we first investigated the role of charge
complementarity. We noted that under the given conditions, the 1 units bear a single positive charge,
whereas the 4 units have two negative charges. Consequently, in oligomers containing the two building
blocks, in a 2:1 (or 4:2) molar ratio, charge complementarity is satisfied, which might contribute to the
stability of this macrocycle. It should be noted that charge neutrality as an organizational principle has
not been considered in any of our previous studies on disulfide DCLs, as these always involved species
168
with exclusively positively charged,
[35,36,38]exclusively negatively charged
[43–49]or neutral building
blocks.
[41,50–52]In order to test this hypothesis, we first prepared a DCL at the same stoichiometric conditions ([1] +
[4] = 2.0 mM, [1]:[4] = 2:1), but in slightly acidic medium (50 mM acetate buffer, pH = 3.2). At this pH,
besides the NH
2groups of the lysine side chains (pK
a2= 8.9) the terminal lysine carboxylayes (pK
a1=
2.2) are also partially protonated. Moreover, the aromatic carboxylates of building block 4 are presumably
also mainly in neutral form. (Although the pKa values for 4 are not known, for the structurally related
thiosalicylic acid pKa = 3.5,
[53]whereas for terephthalic acid pKa1 = 3.5, pKa2 = 4.8.
[.54]) In line with our
expectations, the hexamer 1
44
2is only a minor product under these conditions (Figure 5. 1F), proving
that the specific hexamer can only be preferentially forms preferentially when its net charge is 0.
Figure 5. 3. Elucidating the molecular structure of the mixed hexamer 1442: ESI-Orbitrap mass spectrum of 1442 after mass
selection of the [1442]3+ (m/z = 1083 Da) ion at A) 0 V B) 15 V collision voltage. Inset: three possible positional isomers of
1442.The appearance of [14]3+ indicates the presence of mainly the ortho isomer.
An additional hint pointing towards the role of charge complementarity came from the observation
that in DCLs containing 8-20 mol % 4 (Figure 5. 1B and Figure S5. 58E-I) substantial amounts (up to 35
%) of large mixed macrocycles (1
n4
1and 1
n4
2, 8 ≤ n ≤ 13) were observed. In this stoichiometric range,
169
neither 1
8nor 1
44
2forms, most likely due to entropic reasons. On the other hand, the large macrocycles
are only observed under these conditions. We hypothesized that the oligomers of 1 can incorporate
molecules of 4 via thiol-disulfide exchange. Consequently, these initially positively charged macrocycles
are brought into close proximity of each other due to the presence of negatively charged 4 units. As a
result, the local concentration of building block 1 increases and eventually surpasses the effective
molarity required for the ring closure of the larger macrocycles (see Chapter 2).
[52]Figure 5. 4. Characterization of the supramolecular structure of the 1442 assemblies: A) cryo-TEM shows the presence of a
dense network of fibres; B) AFM shows densely associated fibers forming linear, bundled (red arrow) and circular (blue arrow) assemblies, which are C) 5-15 μm in length and several hundred nm in width; D) CD spectroscopy shows characteristic peptide β-sheet signatures (at 195 and 205 nm) E) ThT fluorescence assay indicates the presence of peptide β-sheets.
Whereas charge complementarity can account for a specific molar ratio (2:1) of the building blocks
within the preferred macrocycle, it does not provide an explanation for the specific observed ring size.
We attempted to elucidate the structure of the mixed hexamer. More closely, we were interested in the
specific sequence of the building blocks within the macrocycle (Figure 5. 3, inset). As transmission
electron microscopy (TEM) indicated the presence of nanoscale aggregates in DCLs containing the
mixed hexamer (vide infra), we expected that it would yield a poorly resolved NMR spectrum.
Consequently, we utilized tandem mass spectrometry (ESI-Orbitrap) to gain further insight into the
molecular structure. We isolated the [1
44
2]
3+(m/z = 1083 Da) ion in the gas phase (Figure 5. 3A) and
170
performed collision-induced dissociation at different collision voltages. At low collision voltages (<15 V),
no fragmentation was observed. At 15 V (Figure 5. 3B), 3 fragments were detected: [1
44
1]
3+(m/z = 1007
Da), [1
4]
3+(m/z = 931 Da) and [1
24
1]
2+(m/z = 812 Da). Whereas 1
44
1and 1
24
1can be fragmentation
products of all possible positional isomers, 1
4can be only produced from the ortho isomer, because only
this macrocycle contains the four 1-units covalently linked to each other (see Figure S5. 59 for details).
Further increasing the collision voltage produced fragments arising from extensive water loss, as well as
b
5fragments (i.e. loss of the C-terminal lysine) of one peptide building block within the macrocycle (Figure
S5. 60). Although unexpected, water loss is a well-described phenomenon occurring during the
gas-phase fragmentation of serine-containing peptide cations.
[55]Notably, the presence of non-specific
fragments does not allow us to exclude the initial presence of the other positional isomers. However, the
ion intensity of the diagnostic fragment 1
4is the highest at the given conditions (compared to 1
34
1and
1
24
2, see Figure S5. 61), which suggests specific formation of the ortho isomer.
Figure 5. 5. Structural organization of the the 1442 assemblies: negative-staining TEM images recorded in a DCL prepared by
mixing 1 and 4 ([1]+[4] = 2.0 mM, [1]:[4] = 2:1) after A) 21 hours, B) 2 days, C) 29 days of DCL preparation. SAXS profiles of the same samples recorded after D) 2 days, E) 34 days after DCL preparation.
After elucidating the molecular structure of the hexamer, we set out to gain insight into its
supramolecular organization. Cryogenic transmission electron microscopy (cryo-TEM) of a 2 days old
171
sample (containing exclusively 1
44
2) indicated the presence of a highly interconnected network of
nanoscale fibers (Figure 5. 4A). However, the length of the fibers could not be determined using
cryo-TEM, as it exceeded the size of the TEM grids. Atomic force microscopy (AFM) was able to capture the
dimension of the fibrous assemblies, showing the presence of thick fiber bundles, with lengths of 5 -15
μm (Figure 5. 4B, C). The morphology of these bundles was diverse: linear, curled up and circular
bundles were likewise observed. In comparison, the fibers resulting from the peptide-only replicators (i.e.
1
8, see Figure S5. 63) are shorter (0.5-5 μm) and only 2-5 nm wide.
[36,56,57]These results indicate that
the fibers are highly interconnected (see also Figure S5. 64G). Moreover, they exhibit a (partial) β -sheet
structure, as indicated by CD spectroscopy (Figure 5. 4D), featuring bands characteristic of parallel
β-sheets (at 208 and 260 nm). Comparing with the CD spectrum of 1
8, these bands are significantly broader
than in the 1-only peptide, whereas a new band appears at 320 nm (Figure S5. 62A). Thioflavin T
fluorescence assay showed the presence of β-sheet structural units (Figure 5. 4E), albeit the mixed
hexamer 1
44
2showed a lower fluorescent response than the known 1
8(Figure S5. 62B). Concerning
the high abundance of the fibers, this result is not necessarily the consequence of a lower beta-sheet
content compared to 1
8, as indicated by the similar intensities in the CD spectra (Figure S5. 62A). The
lower fluorescence might be a result of the strongly associated nature of the fibers, which renders the
peptide beta-sheets less accessible for the thioflavin T dye molecules to intercalate.
We also observed that the originally clear solution of the mixed hexamer 1
44
2turned gradually turbid
upon stirring for several days, whereas the relative amount of the mixed hexamer remained constant in
this period as judged by UPLC analysis. We therefore investigated the temporal evolution of the DCL
using TEM. After 1 day, a wide variety of bundled fiber morphologies was observed, including linear,
circular and curled aggregates, similar to those detected by AFM (Figure 5. 5A). After 2 days, the fibers
adopted a more ordered superstructure, composed of several hundreds of fibers, resulting in 400-600
nm wide fiber bundles (Figure 5. 5B). Remarkably, after 29 days, the initially thick fiber bundles were
transformed into thinner ones, featuring a thickness of 100-200 nm (Figure 5. 5C). However,
negative-staining TEM by its very nature cannot yield a statistically reliable representation of the nanostructures
present in a sample. Moreover, it often suffers from aggregation artefacts resulting e.g. from drying
effects.
[58]In contrast small-angle x-ray scattering (SAXS) could yield direct information on the
dimension(s) and shape(s) of the nanoscale objects as present in their original environment (solution,
suspension etc.). More precisely, the shape of the aggregates can be derived from the slope of the SAXS
profile in a given scattering vector region, whereas their size can be easily calculated from the
characteristic peaks detected. The SAXS profile of a 2 days old sample showed a q
-1slope in the 0.1
nm
-1< q
-1<1 nm
-1region indicating the presence of rod-like objects with a diameter of 2.0 nm (Figure 5.
172
sample (34 days, similar to those shown in Figure 5. 5C), shows several peaks in the 1 nm
-1< q
-1<10
nm
-1region, indicating that during aging the supramolecular structure of the aggregates became more
organized (crystalline). The most intense peak corresponds to a distance of 5.8 nm (Figure 5. 5E), which
most probably originates from scattering on densely packed neighboring fibers, although simple MM
calculations resulted in a distance of 3.6 nm for two neighboring macrocycles. However, more
importantly, SAXS clearly indicates the morphology change towards a more ordered state upon aging,
which cannot be detected in a straightforward manner by mere visual inspection of the corresponding
TEM images (Figure 5. 5B, C).
Given the remarkably high propensity of the mixed hexamer to self-assemble, we proceeded to study
the kinetics of its formation. More specifically, we were interested whether the hexamer is capable of
self-replication, i.e. catalyzing its own formation from the required building blocks. Autocatalytic
processes are characterized by an initial lag phase, in which the (autocatalytic) product is formed via a
non-catalytic pathway from its building blocks, followed by a sudden increase in reaction rate due the
onset of the autocatalytic pathway. In the case of the peptide replicators studied previously, the
aforementioned breakage-growth mechanism confers mechanosensitivity to the system, i.e. the
formation of the replicating species from its building block only occurs upon agitation (stirring) on a time
scale of 15-30 days.
[36]In sharp contrast, the time evolution of the mixed hexamer 1
4
4
2from its monomers
upon agitation ([1] + [4] = 2.0 mM, [1] :[4] = 2:1) featured a very short lag phase of ca. 1 hour, which was
highly variable across replicas, due to the stochastic nature of nuclei formation (Figure S5. 68A,B). The
lag phase could be increased upon decreasing the rate of stirring (Figure 5. 6A) and/or the total building
block concentration (Figure S5. 68C, D). Similarly, the time required to convert 50 mol% of the library
material into 1
44
2required only 3-6 hours instead of several days (as observed for the pure peptide
replicators). Interestingly, the accelerated formation 1
44
2is accompanied with rapid (ca. 1 day, Figure
5. 6A) oxidation of monomers, which takes 5-10 days in the case of the pure peptide self-replicators.
These results indicate that the formation of 1
44
2from its building blocks proceeds with a surprisingly low
kinetic barrier. The underlying reason is possibly related to the net charge of the assembling species:
whereas the pure peptide macrocycles bear a net positive charge under the applied conditions (e.g.
nominally +8 for 1
8), imposing a substantial kinetic barrier on the assembly, 1
44
2is neutral and thus
stacking do not suffer from the barrier resulting from charge repulsion.
The high preference of the system toward the formation of 1
44
2prompts the question whether 1
44
2is
stable as the free molecule (thermodynamic product) or only stabilized upon incorporation into the fibrous
assemblies (kinetic product). In order to distinguish between these scenarios, we simultaneously
monitored the composition and the supramolecular chirality in a stirred DCL, containing the building
blocks in a 2:1 molar ratio (Figure 5. 6B). We observed no lag phase between the onset of hexamer
173
formation and that of fiber formation, i.e. free hexamer molecules could not be detected. This finding
rules out that the hexamer is a stable thermodynamic product on its own and shows that it is stabilized
upon assembly into fibers. We also attempted to get insight into the kinetics of the assembly formation
by monitoring the process with simultaneously with UPLC and TEM. In line with the previous results, as
soon as hexamers were detected by UPLC, TEM showed the presence of various fibrous aggregates.
Moreover, these measurements showed the appearance of a large variety of different assemblies
composed of laterally associated fibers (Figure S5. 64). As we could not draw a clear conclusion from
the juxtaposition of TEM snapshots, we decided to follow the assembly process with real-time AFM
imaging. These measurements are ongoing.
Figure 5. 6.Time evolution of the formation of the mixed hexamer 1442. A) Concentration profile as a function of time in a
non-agitated DCL composed of 1 and 4 ([1]:[4] = 2:1, [1] + [4] = 2.0 mM). B) Time evolution of an non-agitated DCL composed of 1 and
4 ([1]:[4] = 2:1, [1] + [4] = 2.0 mM), showing concomitant increase of the amount of 1442 molecules and the supramolecular CD
signal. C) Seeding experiment in a agitated DCL composed of 1 and 4 ([1]:[4] = 2:1, [1] + [4] = 50 μM) comparing a non-seeded and a non-seeded (15 mol% pre-formed and sheared 1442) sample.
174
In order to prove that the formation of 1
44
2is autocatalytic, seeding experiments were performed: a
small amount of pre-formed assemblies of hexamer (seed) was added to a mixture of the constituent
building blocks (food) and the rate enhancement of hexamer formation, compared to a blank
(non-seeded) sample was measured. However, we faced several difficulties upon performing these
experiments. First, the rate of the non-catalyzed hexamer formation (non-seeded control) was high even
when starting from monomers (vide supra); using oxidized material as food resulted in 61% yield of the
hexamer within 50 minutes (Figure S5. 70). Thus, we used a diluted (50 μM in total) mixture of monomers
as food. This, however, resulted in partial reduction of the seed at the beginning of the experiment, which
resulted in poorly reproducible initial conditions. Second, as the catalytic moieties of the seed are
presumably the fiber ends, we expected the long (and strongly aggregated) hexamer fibers to show low
catalytic activity. According to our previous studies on peptide replicators, we observed that applying
high shear stress (in a Couette cell) on replicator fibers produces a seed with lower (and more uniform)
fiber length, thereby increasing the relative (molar) catalytic activity of the seed. Thus, applying high
shear stress on the seed improves its autocatalytic efficiency.
[56]However, TEM showed that upon
shearing the 1
44
2assemblies, the degree of lateral association decreases, however, the fibers are still
relatively long. Therefore, we performed the seeding experiments with seeds exposed to shear stress
for elongated times (1 hour instead of 30 minutes). The seeding experiment (Figure 5. 6C) showed initial
reduction of the seed; however, after this period, a clear increase in the rate of hexamer formation was
observed, when compared to the non-seeded control. These results confirm the autocatalytic properties
of 1
44
2. It should be noted that the rate difference between seeded and not-seeded samples was not
always observed (Figure S5. 69). Especially, in cases where the lag phase was very short or absent, no
rate difference could be observed. This is, however, not unexpected, given that the rate of formation of
1
44
2is comparable or even higher than that of thiol oxidation (Figure 5. 6A), rendering thus the latter
reaction rate limiting. The oxidation rate is, however, is not affected by the rate of subsequent reactions
(i.e. assembly) and thus is expected to be the same both for seeded and non-seeded samples.
1
44
2is the first example of an artificial multicomponent self-replicator with a specific stoichiometry
and building block sequence (formed spontaneously from its building blocks). We hypothesize that this
specificity results from the interplay of several factors. First, the formation of a neutral species renders
the 2:1 stoichiometry favorable. For entropic reasons, however, larger macrocycles (e.g. 1
64
3) with this
stoichiometry are only formed in minor amounts. Second, the previously observed strong secondary
interactions between primary ammonium ions of 1 and the carboxylates of 4-only oligomers favor the
formation of macrocycles with two or more neighboring units of 4. Thus, the formation of a hexamer 1
44
2is thermodynamically more favorable than that of the trimer 1
24
1, as the latter cannot contain two vicinal
4 molecules. In line with this hypothesis, a DCL prepared from 4 and a neutral building block with the
175
same backbone as 1 (see Chapter 3, Figure 3.1)
[41]features the corresponding trimer as a major and the
hexamer (as a mixture of regioisomers) as a minor product. Moreover, the resulting hexamer 1
44
2is
indeed most probably the ortho isomer, as indicated by MS/MS analysis. Consequently, it possesses a
4-rich (negatively charged, acceptor) and a 1-rich (positively charged, donor) side. Third, the hexamer
1
44
2features a large aromatic core, enabling the stacking of the macrocycles, conferred by the
hydrophobic effect. The characteristic β-sheet signature (Figure 5. 4D) of the stacks indicates that a
large portion of the peptide chains are arranged on top of each other. Consequently, it is highly likely
that within the fibers, several units of two neighboring 4 molecules are also positioned on top of each
other. This creates hydrophobic microenvironments with four carboxylates in close proximity (two from
the upper rim of the bottom 4-dyad and another two from the lower rim of the top 4-dyad). Finally, due to
the presence of these negatively charged 4-dyad spots on the surface of the hexamer fibers, the fibers
have a high propensity to aggregate laterally, as indicated by SAXS experiments. In
nucleation-elongation processes resulting in fibril formation
[59,60](e.g. in the case of dynamic covalent peptide
replicators
[39]), the formation of the primary nuclei proceeds with a high kinetic barrier, mainly because
the number of recognition sites required for the assembly are only available in one specific direction (
i.e. that of the fibril axis). In contrast, in case of 1
44
2, the 4-dyad spots on the surface of one stack can
strongly bind to lysine residues on the surface of another stack (Scheme 5. 1, middle), thereby
accelerating the self-assembly process. This molecular feature explains the high aggregation propensity,
as well as the substantially reduced lag phase and increased rate of self-assembly (compared to the
peptide-only replicators.
5.2.3. Structure and assembly of the mixed tetramer 1
14
3After elucidating the self-assembly properties of 1
44
2, we examined the structure and self-assembly
mode of the mixed tetramer 1
14
3. Given its highly selective and high-yielding (80% UPLC yield) formation
and the presence of complementary binding motifs from the two constituent building blocks, we
hypothesized that it is stabilized by additional intermolecular interactions. However, in contrast to the
mixed hexamer, DCLs containing 1
14
3as a main product (with the same building block concentration)
did not show the presence of aggregates in TEM (Figure S5. 72). Additionally, detailed examination of
the kinetics of its formation from its monomers ([1]+[4] = 2.0 mM , [1]:[4] = 3:1) did not exhibit a lag phase,
even at lower concentrations (Figure S5. 71). We therefore hypothesized that instead of a
supramolecular polymer, it self-assembles into a discrete supramolecular oligomer in form of a cyclic
daisy chain or a lasso [1] pseudorotaxane structure via constitutional self-selection.
[61]176
Figure 5. 7. 1H-NMR spectrum of A) 1 B) (1
143)3 highlighting large upfield shifts of the protons of the lysine side chain and
considerable downfield shifts in the serine and leucine amide protons. C) 1H-1H NOESY spectrum of 1143, highlighting nOe
cross-peaks between aromatic and upfield shifted (threaded lysine side chain) protons (green squares) D) DOSY spectrum of (1143)3, confirming the presence of one species with a diffusion constant D = 2.1×10-6 cm2 s-1 (600 MHz, D2O/phosphate buffer,
pH = 7.2).
Molecular daisy chains
[62]are linear or cyclic supramolecules formed from the non-covalent
oligomerization a covalent conjugate, composed of a thread (donor) and a macrocycle (acceptor). They
177
are potential candidates for various molecular machines, e.g. for molecular muscles. However, they have
been investigated less frequently, due to the lengthy covalent syntheses of the thread-macrocycle
conjugate and its non-selective oligomerization toward daisy chains.
[63]Therefore, a synthetically less
challenging and higher-yielding procedure for the synthesis of daisy chains is required.Surprisingly, 1
14
3showed a clear and well-resolved
1H-NMR spectrum. Therefore, the structure and the assembly state of
the mixed tetramer 1
14
3were elucidated with one-and two-dimensional NMR techniques (TOCSY,
NOESY, dqfCOSY, HSQC, DOSY) by assigning the protons in the parent peptide building block 1 as
well as the tetramer 1
14
3. The full details of NMR assignment are described in the Supplementary
Information, Section 5.4.9 and 5.4.10. In both molecules, the amide protons were identified first based
on the number of their TOCSY cross-peaks; the α-protons were thereafter identified via a series of
NOESY cross-correlations. This was followed by the assignment of the protons of the single amino acid
side chains by TOCSY cross-correlations with the corresponding amide protons; overlapping signals
could be resolved based on their
1H-
13C HSQC cross-correlations.
First, we investigated the structure of the peptide building block 1. The triplet signals at 2.91 and 2.97
ppm were assigned to the K3ε and K5ε methylene protons, respectively, whereas the broad multiplet at
1.4 ppm could be assigned to the K3γ and K5γ side chain protons (Figure 5. 7A). In sharp contrast, the
mixed tetramer 1
14
3features only 1 peak in the region expected for Kε protons, whereby no peak
appeared in the region corresponding to Kγ protons (Figure 5. 7B). However, new signals were detected
in the upfield region (+0.7- -1.65 ppm). These could be assigned to the K3 side chain protons (by their
TOCSY cross-peaks with K3α), which accounts for large upfield shifts (up to -3 ppm in case of K3 ε),
indicating considerable shielding of a diamagnetic ring-current. Additionally, K3α shifted strongly upfield
(by -1.4 ppm), whereas G1α and S4α protons underwent a strong downfield shift (+0.6 ppm and +1.7
ppm, respectively). Additionally, nOe cross-peaks were detected between the protons of the K3 side
chain and the aromatic protons of the macrocycle (Figure 5. 7C). These observations strongly support
the presence of a noncovalent oligomer of 1
14
3, featuring the K3 lysine side chain threaded through the
aromatic macrocycle. This topology also accounts for the considerable shielding of the K3 protons as
they experience a large diamagnetic ring current inside the macrocycle.
Furthermore, in 1
14
3the K5 ammonium group gives a relatively sharp
1H-NMR peak at 7.50 ppm,
whereas no such peak is detected for 1. Thus, the ammonium protons exchange considerably slower in
1
14
3as in 1, which possibly results from their strong electrostatic binding to the carboxylate groups of
the macrocycle. Similar NMR signatures were detected previously for the structurally analogous 4
4.Sp
pseudorotaxane, also featuring an alkylammonium chain complexed by a structurally similar
macrocycle.
[43]The threaded complex is stabilized by the electrostatic interaction between the
ammonium group of K3 and the carboxylates of the macrocycle as well as by hydrophobic self-sorting
178
(threading of hydrophobic alkyl chains through the hydrophobic aromatic cavity). DOSY experiments
indicate the presence of a single species with a diffusion coefficient of 2.1×10
-6cm
2s
-1(Figure 5. 7D).
Figure 5. 8. A) Negative mode ESI-Orbitrap spectrum of 1143, showing the presence of [c3] daisy chain (1143)3 as well as dimer
fragment (1143)2 and monomeric 1143. B) SDS-PAGE analysis of 1143, showing the presence of only one species, with a mass
slightly higher than 3.5 kDa, thus corresponding to [c3] daisy chain (1143)3. B) SDS-Page analysis of the [c3] daisy chain formed
from 1143 ([1]+[4] =2.0 mM, [1]:[4] = 1:3.5). Lanes: 1: ultra-low weight molecular ladder, 2: 1143 in phosphate buffer (50 mM, pH
= 8.2), 3: 1143 in ammonium acetate buffer (50 mM, pH = 7.0).
Next, we set out to determine the exact composition of the cyclic daisy chain formed from 1
14
3.
Negative mode mass spectrometric analysis (ESI-Orbitrap, direct injection) of a 2.0 mM sample prepared
in ammonium acetate buffer (pH = 8.2, 50 mM) showed the presence of the monomer ([1
14
3]
-, 690 Da)
as well as the noncovalent dimer ([(1
14
3)
2]
3-, 920 Da) and trimer ([(1
14
3)
3]
4-, 1035 Da) in the gas phase
(Figure 5. 8A). Notably, the peak at 1380 Da shows that monomer, as well as dimer and trimer are
present (indicated with cyan circles, rectangles and triangles, respectively). Regarding that the mildest
ionization conditions possible were applied, we tentatively assign the peaks corresponding to the
monomers and dimers to fragment ions and those of the trimer to the parent ion, i.e. to a [c3] daisy chain
(1
14
3)
3. The low stability of the [c3] daisy chain can be explained by the fact that a significant portion the
non-covalent binding energy in solution is attributable to the hydrophobic self-sorting effect, which
vanishes upon transferring the molecules to the gas phase.
[64]Accordingly, no higher oligomers (i.e.
[(1
14
3)
4]
5-, 1840 Da) were detected in the gas phase, suggesting that the oligomerization is
non-statistical. This assumption is corroborated by the results of SDS-PAGE analysis (Figure 5. 8B), which
indicated the presence of one species in solution with a molecular mass slightly higher than 3.5 kDa,
179
corresponding to the [c3]-daisy chain (4.2 kDa). Notably, the daisy chain is stable under denaturing
conditions, as no bands corresponding to the lower oligomers were detected.
Our next goal was to assess the feasibility of threading by molecular dynamics (MD) simulations.
Specifically, we considered the self-threaded macrocycle in form of a pseudo[1]rotaxane
[65,66]as well as
oligomeric closed (cyclic [cN] daisy-chain, with all K3 lysine side chains being threaded) and oligomeric
open (acyclic [aN] daisy-chain, with unthreaded terminal K3 lysine side chains) assemblies of 1
14
3.
[67]Results of the MD simulations are summarized in Table 5. 1. First, we found that a pseudo[1]rotaxane
is energetically uphill: strain induced by the bending of the peptide chain is not entirely compensated by
favourable interactions between the lysine side chain and the macrocylic cavity and results in an
enthalpic penalty of +8 kJ/mol compared to the non-threaded monomer.
Bond strain could be lower in the case of intermolecular threading, thus, we next considered the
smallest oligomeric assembly, i.e. the threaded dimer (N=2). The simulations indicated that the formation
of the acyclic [a2] daisy chain does not come at the cost of bond strain, resulting in the non-threaded
conformer being enthalpically slightly favorable (ΔH = -5 kJ/mol) compared to a threaded, but
non-covalently aggregated state. However, further intracomplex threading towards a cyclic [c2] daisy chain
did not lead to a further enthalpic gain due to the enhanced bond strain as a consequence of looping
around the second peptide chain.
We hypothesized that the next larger cyclic non-covalent oligomer, i.e. the cyclic [c3] daisy chain
would not suffer from internal strain any more, considering the extra conformational freedom. MD
simulations of the [a3] and [c3] daisy chains were performed, but surprisingly showed a high enthalpic
cost of threading (=20 kJ/mol per macrocycle). However, it should be noted that due to the increased
size and flexibility of the complex, MD simulations have been so far unable to sample its whole free
energy landscape in a reasonable time scale, making it less likely that the lowest energy state was found
for this complex. It should also be noted that the simulations outlined above do not take into account the
entropy associated with the threading.
Most notably, Entry 1 in Table 5. 1 (monomer) shows that the macrocycle has to “breathe open” to
allow the lysine side chain to enter. No spontaneous threading or dethreading of 1
14
3molecules was
observed in any of the simulations. This finding suggests that the kinetic barrier for the (de)threading of
single tetramer molecules is large and that the threading mainly proceeds by covalent formation of the
macrocycle around the lysine “template”. We therefore concluded that the formation of the 1
14
3macrocycles (covalent bond formation) and the threading event (non-covalent bond formation) occur
simultaneously, i.e. the daisy chain is formed via subcomponent assembly rather than simple
self-assembly.
180
Table 5. 1. Relative enthalpies of threading from MD simulation results. aIndividual contributions in bonded and non-bonded
energy, respectively, are given in brackets and per monomer. Unthreaded simulations with the same number of macrocycles, water and ions were used as a reference. bSnapshots exclude water, ions and non-N hydrogen atoms for clarity.
System
ΔH
thread(kJ/mol/macrocycle)
aSimulations snapshot
b
Unthreaded
single
macrocycle
2298 Water
9 Na
+4 Cl
-(0)
Self-threaded
single
macrocycle
2298 Water
9 Na+
4 Cl
-+8 (19-12)
vs. unthreaded
single macrocycle
Single threaded
dimer of
macrocycles
2358 Water
17 Na+
7 Cl
--5 (0--5)
vs. unthreaded
dimer
181
Double
threaded dimer
of macrocycles
2358 Water
17 Na+
7 Cl
--3 (12-15)
vs. unthreaded
dimer
Triple threaded
trimer of
macrocycles
2960 Water
24 Na+
9 Cl
-+20 (9+11)
vs. unthreaded
trimer
In simple self-assembly of supramolecular architectures, the components (e.g. metal and ligand or
axle and macrocycle), previously synthesized by covalent chemistry spontaneously assemble via
non-covalent bonds. This methodology is, however, somewhat prohibitive as the non-covalent synthesis of the
components could be rather lengthy and low-yielding, as mentioned before. In contrast, in subcomponent
self-assembled complexes the two different type of bonds (covalent and non-covalent) are formed
simultaneously, e.g. chelating ligands are produced via reversible (imine) bond formation while at the
same time metal-ligand bonds are constructed.
[68]This technique has indeed frequently been used in the
construction of large metallosupramolecular architectures.
[69]However, as far as we are aware,
subcomponent self-assembly of cyclic daisy chains have never been described. Our analysis proves the
first instance of the formation of a [c3] daisy chain via subcomponent self-assembly, under mild
conditions and in aqueous solution, resulting in a high-yielding synthesis of the final supramolecular
architecture from a dynamic combinatorial library.
182
5.2.4. Relative stabilities of the self-assembling species in the (1+4)-system
Next, we investigated the relative thermodynamic stabilities of the three self-assembling macrocycles
(1
8, 1
44
2, 1
14
3) in the system composed of 1 and 4. As seen from our previous results, the assemblies of
these three molecules are held together by remarkably different patterns of supramolecular interactions:
1
8is organized by the stacking of the aromatic cores and the β-sheet interactions between the peptide
side chains. 1
44
2is additionally stabilized by multivalent ionic interactions between the lysine side chains
and the core carboxylates, causing extensive internal association of the fibers. In sharp contrast, 1
14
3is
only stabilized by three lysine-carboxylate moiety and the hydrophobic effect resulting from the threading
of the alkyl chain of lysine through the aromatic core. Additionally, we considered 4
4.Sp, as it is held
together by solely one alkyl moiety (host) threaded through a ring (guest). To study the relative stabilities
of the assemblies, we reacted the DCLs containing one assembly (1
8, 1
44
2or 1
14
3) with 1, 4 or Sp
(Scheme 5. 2). At the start of the reactions disulfides were partially reduced (20 mol%) in order to ensure
that thiol-disulfide exchange could take place. The amount of monomers (or Sp) were calculated to
achieve a theoretical 1:1 molar ratio of substrate and product (as indicated in the equations shown in
Figure 5. 9). Any substantial deviation from the expected ratios would indicate a thermodynamic or
kinetic preference towards the species formed in a higher amount than predicted.
Scheme 5. 2. Overview of selected interconversion processes between self-assembled species 18, 1442, 1143, 44.Sp via addition
183
Figure 5. 9. Comparison of initial (green, empty bars), expected final (red, striped bars) and measured (blue, filled bars) relative
UPLC intensities in the interconversion processes between self-assembled species 18, 1442, 1143, 44.Sp. All DCLs are prepared
in phosphate buffer (50 mM, pH = 8.16) stirred at 1200 rpm and contain 20 mol% thiolate at the beginning. A) 18 and 4 (0.67
eq.) affords 1442. B) 1442 and Sp (32 eq.) fails to produce 18. C) 1442 and 4 (2.0 eq.) affords a 5.2:1 mixture of 1143 and 1442. D)
1143 and 1 (2.0 eq.) results in somewhat lower than expected conversion to 1442. E) 1143 and Sp (0.5 eq.) results in the expected
conversion to 1442 and 44.Sp. F) 44.Sp and 1 (0.57 eq.) fails to produce 1143 and yields 1442 as a major product instead. G)
44.Sp and 1 (2.7 eq.) affords 1442 in the expected yield. H) 18 and 4 (0.5 eq., 33 mol % thiolate) fails to produce 1143 and affords
1442 instead. I) 1143 and 1 (0.75 eq., stoichiometric amount) fails to give 18 and affords 1442 instead.
First, the interconversion between 1
8and 1
44
2was examined. In the reaction of 1
8and partially
oxidized 4 the mixed hexamer 1
44
2was obtained as the main product, although in a lower amount than
expected. This deviation is presumably due to the stabilization of the 1
8molecules in the fibrous
aggregates (Figure 5. 9A). In contrast, attempts to sequester building block 4 from the mixed hexamer
by addition of even a large excess (32 equivalents) of Sp and to subsequently obtain 1
8, failed (Figure
5. 9B). We attribute the high stability of 1
44
2towards Sp to the high number and density of strongly
interacting lysine-carboxylate moieties within the hexamer assemblies. This assumption is supported by
the feasibility of such an effector-triggered self-replication process in a different DCL obtained from 4
184
and an analogue of 1. This analogue bears an oligo(ethylene-oxide) chain, which only moderately
engages in intermolecular interactions, thus enabling the sequestering of 4.
[41]Second, we investigated
the mutual interconversion of 1
44
2and 1
14
3. Addition of 2 equivalents of 4 to 1
44
2affords a 5.2:1 mixture
of the hexamer and tetramer (Figure 5. 9C). The large deviation from the calculated 1:1 ratio possibly
originates from the high relative stability of the hexamer (the rest of 1 is incorporated in various smaller
oligomers). In contrast, the conversion of 1
14
3to 1
44
2either by the addition of 2 equivalents of Sp (Figure
5. 9D) or 2 equivalents of 1 (Figure 5. 9E) gives rise to the mixed hexamer in the expected yields. These
results indicate that the conversion of 1
14
3to 1
44
2is thermodynamically as well as kinetically favored
compared to the reverse reaction. Third, we investigated the conversion of 4
4.Sp to either 1
14
3(Figure
5. 9F) or 1
44
2(Figure 5. 9G) by addition of 1 (0.57 and 2.7 equivalents, respectively). We expected both
mixed macrocycles to be more stable than the original host-guest complex, due to the multivalent
interactions holding together the assemblies. Surprisingly, in the former case, the reaction failed to yield
1
14
3as a main product and a mixture of 1
44
2and 4
4.Sp was obtained, featuring only trace amounts of
1
14
3(Figure 5. 9F). In other words, if given a chance, the system prefers to form 1
44
2, even if the
stoichiometry would dictate otherwise. Previously, simulations on thermodynamically governed DCLs
have shown that if DCL members do not interact, then the formation of a large number of smaller
macrocycles is preferred to the formation of a lower number of larger ones.
[70–72]In this case, however,
this trend is reversed, which again points out that the formation of 1
44
2is thermodynamically highly
preferred, due to the enthalpy gain arising from the interactions between lysine side chains and the
neighboring carboxylates. In contrast, upon addition of higher amounts of 1 (Figure 5. 9G) the expected
conversion to 1
44
2was observed (although the amount of 4
4.Sp was lower than expected, due to
incorporation of 4 into minor products). Finally, we investigated the interconversion between 1
14
3and 1
8.
Upon addition of 4 to 1
8, instead of the expected 1:1 mixture of 1
14
3and 1
8, 1
44
2emerged as the main
product, alongside minor amounts of residual 1
8, but most importantly, without even traces of 1
14
3(Figure 5. 9H). This outcome is, however, in line with the previous interconversion reactions and once
again points out the high thermodynamic preference towards 1
44
2, which renders the otherwise fast and
high-yielding formation of 1
14
3kinetically unfavored. Similarly, based on the quantitative conversion of
1
14
3into 1
44
2, as well as the reluctance of the latter to give 1
8upon addition of Sp, it is not surprising that
addition of large excess of Sp to 1
14
3(Figure 5. 9I) failed to produce the octamer 1
8and yielded the
hexamer 1
44
2. Overall, these experiments point out that A) 1
44
2acts as a thermodynamic sink, i.e. the
presence of multiple binding sites in the aggregates completely blocks its guest-induced conversion into
other species B) 1
8is stabilized to a lesser extent and can be converted to other species, although with
lower yields, 3) 1
14
3and 4
4.Sp are significantly less favored, and can be easily converted to other species
185
5.2.5. Substrate scope
Finally, we investigated the possible formation of analogous macrocycles from structurally related
building blocks (Figure 5. 10). First, peptide building blocks 2 and 3, containing an alanine or threonine
residue at position 5 (instead of serine), respectively, were mixed with 4 in molar ratios 2:1 and 1:3.5.
After no change in composition was detected, UPLC and LC-MS analyses of the resulting DCLs indicated
analogous formation of mixed hexamers (2
44
2, 3
44
2) and tetramers (2
14
3, 3
14
3) for the 2:1 and 1:3.5
mixtures, respectively (Figure 5. 10A-D). Characterization of the mixed hexamers (2
44
2, 3
44
2) with CD
spectroscopy (Figure S5. 87A) and ThT fluorescence assay (Figure S5. 87B) indicated the formation of
β-sheet assemblies. In the case of both hexamers, TEM analyses showed the presence of several
micrometer long fibrous assemblies similar to those of 1
44
2. These results show that upon small
modifications at the molecular level the properties at the systems level (dual mode of self-assembly) are
preserved.
Next, we examined how a deeper modification of the peptide building block influences the systems
properties. Peptide 5, bearing a phenylalanine moiety at position 5, was mixed with building block 4 in
2:1 and 1:4 molar ratios, respectively, under identical conditions as applied previously. Remarkably, the
selective formation of the analogous 5
44
2hexamer was not observed in the former case (Figure 5. 10E);
however, the tetramer 5
14
3was formed almost exclusively in the latter system (Figure 5. 10F).
Presumably, the bulky phenyl rings in 5 render the barrier for the orchestrated lateral association of a
high number of hexamer macrocycles (assembled into short stacks) too high. Consequently, in the
absence of stabilizing ionic interactions, there is no thermodynamic preference for the formation of the
hexamers. In contrast, for the formation and threading-mediated stabilization of the tetramer 5
14
3, only
three tetramer molecules have to be accommodated in a suitable position, which is apparently still
possible even in the presence of the bulky phenyl groups in close proximity of the coordinating lysine
residues.
At last, we examined the effect of a small modification of the non-peptidic building block on the
system. DCLs were prepared by mixing building block 1 with the previously synthesized dithiol building
block 6 (2,6-dimercaptoterephthalic acid), which differs from building block 4 solely in the position of one
thiol group. Remarkably, in this case none of the library members formed specifically and a diverse
mixture of isomeric macrocycles were formed (Figure 5. 10G-H).
186
Figure 5. 10. UPLC chromatograms of DCLs prepared from building blocks 2 + 4 (A, B), 3 + 4 (C, D), 5 + 4 (E, F) and 1 + 6 (G,
H) at a total building block concentration of 2.0 mM and building block ratios 2:1 (A, C, E, G) and 1:3 (B, D, F, H) after 3 days (A-F) or 9 days (G-H) of stirring (in 50 mM phosphate buffer, pH = 8.16).