Impact of Positional Isomerism on Pathway Complexity in Aqueous Media

Abstract Pathway complexity has become an important topic in recent years due to its relevance in the optimization of molecular assembly processes, which typically require precise sample preparation protocols. Alternatively, competing aggregation pathways can be controlled by molecular design, which primarily rely on geometrical changes of the building blocks. However, understanding how to control pathway complexity by molecular design remains elusive and new approaches are needed. Herein, we exploit positional isomerism as a new molecular design strategy for pathway control in aqueous self‐assembly. We compare the self‐assembly of two carboxyl‐functionalized amphiphilic BODIPY dyes that solely differ in the relative position of functional groups. Placement of the carboxyl group at the 2‐position enables efficient pairwise H‐bonding interactions into a single thermodynamic species, whereas meso‐substitution induces pathway complexity due to competing hydrophobic and hydrogen bonding interactions. Our results show the importance of positional engineering for pathway control in aqueous self‐assembly.


Introduction
Biomolecular systems are able to adapt their morphology in aqueous media to different environmental conditions. [1] A particularly illustrative example is the a-protein tropomyosin, which can exist in eight distinct types of aggregates. [2] This unparalleled pathway control has motivated the investigation of these phenomena in synthetic self-assembled counterparts. In recent years,p athway complexity [3] has been observed for different types of building blocks both in non-polar [4] and aqueous media. [5] Control over the competing aggregation pathways in these molecular systems is typically achieved by optimization of sample preparation protocols.These processes are generally tedious and often require avery specific set of experimental conditions (that is,concentration, temperature, solvent composition, and others). [6] Alternatively,these complex sample preparation methods could be simplified by molecular design strategies,w hich have been recently introduced to broaden the scope of pathway complexity.T od ate, these approaches have mainly focused on the geometrical modification of the building blocks either by systematic size variation of substituents [4d,h] or length variation of ag iven molecular fragment (for example, p-system, [4g,5m] alkyl spacers, [4e, 5d] or side chains [5a,j] ). In this regard, controlling pathway complexity in aqueous media by molecular design is particularly challenging,a st he competition between hydrophobic and other non-covalent interactions [5b-f,h,i,k,l] makes the selfassembly considerably less predictable.I na mphiphilic selfassembly,f ine-tuning of the hydrophilic/hydrophobicr atio is aw ell-known strategy for morphology control [7] and, to am inor extent, has also been observed to induce pathway complexity. [5a,j,m] However,u nderstanding the molecular design principles that govern the existence of pathway complexity in self-assembly still remains an open question and new approaches are required.
In the present manuscript, we introduce the concept of positional isomerism as an ew molecular design strategy for pathway selection in aqueous self-assembly.T ot his end, we have synthesized two positional isomers of an amphiphilic BODIPY dye that only differ in the location of the functional groups (compounds 1 and 2 in Scheme 1; for synthesis and characterization, see the Supporting Information). In addition to the typical methyl substitution of the BODIPY core, [8] both dyes are functionalized at the 6-position with an ethynylbenzene moiety bearing three hydrophilic triethylene glycol (TEG) chains to provide water solubility.A dditionally,t he choice of ac arboxylic acid group at either the 2-(1)o rt he meso-position (2)ofthe BODIPY core was made based on its potential to establish hydrogen bonds both in non-polar solvents [9] and water when arranged in al ocal hydrophobic environment. [5d, 10] Finally,t he presence of ab ulky aromatic electron-withdrawing substituent at the BODIPY mesoposition is expected to provide the system with enhanced dynamics by hindering af ace-to-face stacking into nonemissive H-type aggregates. [11,12] Thel ocation of the carboxylic acid is observed to determine pathway complexity:i ft his group is placed on the opposite side of the hydrophilic chains,pairwise hydrogen bonding of the carboxyl groups can efficiently occur in ahydrophobic microenvironment, leading to thermodynamically controlled lamellar nanostructures.I nc ontrast, the location of the acid at the meso-position induces competition of entropic and enthalpic contributions:h ydrophobic and aromatic interactions initially drive the self-assembly into metastable unordered nanostructures.O ver time,s teric effects induce am olecular rearrangement into thermodynamically controlled flexible J-type [13] nanofibers that are stabilized by partial hydrogen bonding of the carboxylic acids.

Results and Discussion
First insights into the self-assembly of 1 and 2 were gained from solvent-dependent absorption studies (c = 20 mm, 298 K). Compound 1 exists in amonomeric state (M) in both polar and non-polar organic solvents,s howing absorption maxima at about 577 nm (S 0 !S 1 transition of the BODIPY) along with alow-intensity transition (S 0 !S 2 )atabout 406 nm ( Figure S11, Supporting Information;f or selected "good" solvents such as dichloromethane (DCM) and 1,2-propanediol, see Figure 1a). Similar spectra with characteristic BODIPY bands at 574 nm and 409 nm obtained for derivative 2 (Figure 1b)s how that the different substitution pattern negligibly influences the spectral changes in the molecularly dissolved state.T his is further supported by the presence of sharp emission bands with maxima at comparable wavelengths (l % 620 nm) for both molecules in avariety of organic solvents (Figures 1c,d and S11, Table S1).
This behavior changes considerably if the systems are investigated in water (pH = 7). For 1,t he concomitant decrease in absorption and spectral broadening compared with the spectra recorded in organic solvents are indicative of aggregation ( Figure 1a). Thenegligible shift of the absorption maximum during this process suggests that an ideal face-toface H-type stacking of the BODIPY dyes [14] may possibly be hindered by the sterically demanding benzaldehyde group at the BODIPY meso-position. Emission studies show ar edshift and drastic quenching of the fluorescence (Figure 1c).
For 2,UV/Vis absorption and emission studies of afreshly prepared solution in water show at rend very similar to 1 (green spectra in Figure 1b,d), indicating weakly coupled dye units in the assembled structure (denoted as aggregate A in the following). Surprisingly,u pon aging for am inimum of 20 ho ra fter thermal annealing,a ggregate A transforms into anew,energetically more favorable assembly B with strong Jtype exciton coupling (Dl = 40 nm, l max = 615 nm;F igure 1b, blue spectrum), which is accompanied by acolor change from violet to blue (Figure 1b,i nset). This aggregate species B displays higher emission intensity than A [fluorescence quantum yield, f F (B) = 6.4 %v s. f F (A) = 2.4 %],i na ccordance with ap lausible J-type aggregate formation. Thea dditional blue-shifted emission band at 570 nm may be attributed to the existence of defects in the packing of B (Scheme 2), which is supported by different excitation spectra upon collecting emission at 570 or 650 nm ( Figure S12).
This different propensity of 1 and 2 to undergo pathway complexity prompted us to investigate the nanoscale morphology and size of 1, A,a nd B by cryogenic transmission electron microscopy (cryo-TEM), TEM, and dynamic light scattering (DLS). As shown by TEM and cryo-TEM, compound 1 self-assembles into rigid fibrils with lengths of % 800 nm and au niform diameter of approximately 5nm (Figures 2a-c and S13), which closely matches twice the molecular length. Thei mages also reveal that the individual straight fibrils have astrong tendency to bundle into lamellar structures.These results can be explained by the formation of extended stacks of hydrogen-bonded dimers of 1 (via the carboxylic acid). [5d] Within this arrangement, the polar TEG chains would be exposed to the aqueous medium, thus  enabling the bundling of these fibrils by strong lateral association of the TEG chains,a sa lso observed for other amphiphilic molecules. [15,16] Theanisotropy of these structures is further supported by the observation of angular-dependent hydrodynamic radii in DLS studies ( Figure S14). [17] On the contrary,t he assemblies of A are identified by TEM as anisotropic nanostructures with ill-defined shape ( Figure 2d). Aging this solution leads to the formation of flexible fibres (B)with lengths of % 900 nm and adiameter of 5nm( Figures 2e,f and S13). Compared to 1,t hese fibers are more flexible and have al ower tendencyt ob undle.T he changes in size and morphology for the transformation of A to B were also monitored by DLS studies ( Figure S14).
To gain mechanistic insights into these different aggregation processes,s pectroscopic studies using different ratios of good/poor solvent (THF/water) were carried out at various concentrations (8-40 mm,F igures 3a nd S15-S17). Fort his purpose,the disassembly of aqueous aggregates of 1, A,and B was monitored by the addition of increasing amounts of the respective monomer solutions in THF.Addition of aliquots of monomeric 1 in THF to an aggregate solution of 1 in water at 20 mm initiated the disassembly process,w hich was complete when the volume fraction of THF in water exceeded 50 % (Figures 3a,S 15). Then on-sigmoidal plot of the fraction of aggregated species (a agg )v s. the volume fraction of THF monitored at 572 nm indicates ac ooperative process.S ub-  sequent fit to the denaturation model [18] (Figure 3b)yields an average Gibbs energy (DG8 8)o fÀ52.4 kJ mol À1 (Table S2).
Forc ompound 2,f ull denaturation of the kinetically controlled species A requires alower volume fraction of THF ( % 30 %) than 1 at the same concentration (Figures 3c and S16), indicating ah igher stability of the latter aggregate. Additionally,the disassembly process of A is described by an isodesmic mechanism (Figure 3d), yielding a DG8 8 of À44.3 kJ mol À1 .I nc ontrast, B follows ac ooperative mechanism (Figures 3e,f and S17) and is characterized by am ore negative DG8 8 (À49.3 kJ mol À1 )t han A.A ccordingly,t he assembly of 1 is more favorable than that of A and B,w hich might result from hydrogen bonding of the carboxylic acid groups in the hydrophobic interior of the aggregate.Furthermore,the fact that only the kinetically controlled aggregate A follows the isodesmic mechanism suggests that this pathway is mainly driven by aromatic and hydrophobic interactions,a s typically observed for amphiphilic p-systems that lack directional non-covalent interaction patterns,s uch as hydrogen bonding. [19,17] On the contrary,s uch additional directional interactions are expected to stabilize the assemblies of B and 1 via ac ooperative growth mechanism. [20,15] To unveil the relationship between assemblies A and B, time-dependent UV/Vis experiments at different temperatures and concentrations were performed (Figures 3g,h and S19). As also shown previously by UV/Vis,k inetically controlled species A readily converts to the thermodynamic product B simply by keeping the 20 mm aqueous aggregate solution of A at room temperature over the period of approximately one day (Figure 3g). This transformation can be accelerated by increasing concentration and/or temperature (see plot of Abs 615 vs.time in Figure 3h), which is typical for the consecutive transformation of an on-pathway species. [6] Therefore, A converts into B via ar earrangement of the dye molecules within the aggregate,w hich does not involve the disassembly into free monomeric species.T his transition between two precise aggregate species is also corroborated by ac lear isosbestic point at 584 nm.
Ther esulting sigmoidal plots are diagnostic of an autocatalytic kinetic process, [21] which is in accordance with the faster A!B transformation when A is allowed to coexist with small traces of B in solution ( Figure S19). Also,t he method used to isolate both aggregates starting from molecularly dissolved 2 differs significantly:a ggregate A preferentially forms by rapid cooling (thermal quenching) or fast injection of the monomer solution into al arge volume of water (solvophobic quenching). In contrast, the direct isolation of thermodynamic species B from the molecularly dissolved state is only possible by slow cooling (0.5 Kmin À1 )o fdiluted monomer solutions (c = 8 mm)using a10%volume fraction of THF in water. Unless asmall percentage of THF is added, the aqueous assembly B will not dissociate into the monomer species,n ot even after heating the sample at ah igh temperature (368 K) for ap rolonged time and using low concentrations (2.0 mm). Thea ddition of THF lowers the energy barrier derived from the hydrophobic effect [5j,l] and therefore destabilizes the kinetically controlled species A,w hich is dictated by the hydrophobic collapse.C onsequently,adirect transition from the monomer species to thermodynamic assembly B is facilitated via an ew pathway,w hich does not involve the formation of kinetically controlled state A.
After finding the appropriate experimental conditions [water/THF (9:1), 8 mm,c ooling rate:0 .5 Kmin À1 ], the monomer-to-B transformation was monitored by variabletemperature (VT) absorption and emission spectroscopy upon cooling monomer solutions from 333 Kt o2 83 K. VT UV/Vis studies show the depletion of the monomer band at about 572 nm at the expense of the previously described Jtype aggregate band at 615 nm (Figure 4a), which is the exact reverse behavior observed in denaturation experiments (Figure 3e). Theplot of a agg vs.temperature extracted from these experiments was fitted to the cooperative nucleation-elongation model, [22] yielding a DG8 8 of À31.7 kJ mol À1 (Table S3,   (Figure 4b). Thec hanges in emission at 642 nm as af unction of temperature were also satisfactorily fitted to the nucleationelongation model, yielding thermodynamic parameters that are in good agreement with those obtained from VT absorption studies (Figure 4c,T able S3).
The DG8 8 values for B derived from VT spectroscopic studies are less negative than those extracted from previous denaturation experiments,which is not surprising considering the different experimental conditions used in both experiments:while the starting point for denaturation studies is the aggregate species in pure water, VT studies required the use of 10 %v olume fraction of THF in order to disassemble the aggregates at high temperature,w hich obviously attenuates the overall aggregation tendency of the system. Furthermore, this energy difference can be related to the decrease of entropic contributions with the addition of ac o-solvent, as also recently reported by Würthner and co-workers. [19b] To unravel the enthalpic and entropic contributions to the aggregation processes of the three species 1, A,a nd B, isothermal titration calorimetry (ITC) dilution studies were recorded. Higher concentrations (1 mm)a nd small percentages of THF (10 %) had to be used in order to obtain sufficiently clear heat signals.T he experiments were performed by addition of aliquots (2.5 mL) of the respective aggregate solutions to al arger volume (300 mL) of the pure solvent mixture [water/THF (9:1)].F or all three samples, dilution of the aggregate solution by injection into the solvent mixture water/THF (9:1) produced an endothermic heat flow due to the partial dissociation of the self-assembled structure ( Figure 5). [23] Thea rea of these raw heat signals was plotted against the injection number and fitted to the independent binding site model to determine the thermodynamic parameters of the respective disassembly process (Table S4, Figure S20). Under the investigated experimental conditions,all aggregates (1, A, B)are entropically disfavored and enthalpically favored (Figure 5d). Thee ntropic penalty related to these systems is expressed by the ratio between TDS8 8 and DH 298 (Table S4). Among all aggregates, A exhibits the smallest entropic penalty (0.5) and the highest Gibbs energy (DG 298 = À24.1 kJ mol À1 ), which highlights the key contribution of hydrophobic interactions to the self-assembly process of A.T his penalty increases after the transformation to species B is complete (0.8) and the enthalpy (DH 298 )i s lowered from À50.1 to À110.7 kJ mol À1 ,w hich results in an overall lower Gibbs energy (DG 298 = À25.7 kJ mol À1 ). These much lower enthalpy values for B compared to A support our previous hypothesis that the assembly of B is stabilized by additional interaction patterns compared to A.
On the contrary,the assemblies of 1 are the energetically most favorable ones,a se vident from their more negative DG 298 values (À30.3 kJ mol À1 )c ompared to A and B.F urthermore, 1 exhibits the lowest DH 298 value by far (À286.1 kJ mol À1 ), which is in agreement with the existence of strong directional interactions,that is,pairwise interactions of carboxylic acids in the hydrophobic interior of the nanostructure.Also,the highest entropic penalty (0.9) obtained for this system reflects the high degree of order and rigidity of self-assembled 1.T he thermodynamic parameters obtained by ITC differ from those extracted from previous denaturation studies (for instance,l ess negative DG8 8 values for the former), as also observed when comparing denaturation and VT UV/Vis studies.This discrepancy can be explained by the use of 10 %THF and much higher concentrations (about two orders of magnitude) for ITC compared to denaturation  studies.I nc ontrast, comparable thermodynamic data are obtained by ITC and VT UV/Vis,asthese experiments were performed in the same solvent mixture (10 %THF in water). However,for our particular system, we believe that VT UV/ Visi samore appropriate method to extract the thermodynamic parameters,a sI TC is only able to cover av ery small portion of the disassembly curve due to the much higher concentrations required for this measurement.
To rationalize the molecular packing of the assemblies, four dimers of 2 were optimized using the dispersioncorrected semiempirical PM6 method in the vacuum (Figure S21, Table S5). In the most stable dimer structure,t he BODIPY units are stacked in aslipped or J-type fashion (V = 268 8,F igure S21) with ap arallel orientation of the transition dipole moments [12] (Scheme 2). [24] Thee nergy penalty associated with this parallel dye arrangement can be largely overcome by the formation of hydrogen bonds between the carboxylic acids along with strong p-p interactions between offset-stacked BODIPY units.W hent his calculation is extended to atetramer,itbecomes obvious that acontinuous hydrogen bonding network is not possible due to increased distances between the second and third molecule (3.8 , Figure S22). Thus,t he extended p-stacks with J-type exciton coupling experimentally observed for B are most likely further stabilized by partial intermolecular hydrogen bonding between the carboxylic acid groups,a sp redicted by the calculations (see Scheme 2). Considering that B is not formed directly from the monomer in pure water but rather upon molecular rearrangement of the kinetically controlled species A (see Figure 3g,h), the molecular packing of both aggregates should not differ significantly.O therwise,d isassembly of A into the monomeric species would be unavoidable prior to the formation of B.O nt his basis,aparallel arrangement of the BODIPY transition dipole moments can also be expected for A.Nevertheless,the lack of significant shifts in the absorption spectrum during the formation of A suggests weaker aromatic interactions than for B.M ost likely,s teric repulsion between the bulky benzaldehyde groups at the BODIPY mesoposition during the face-to-face approach of the monomer units prevents strong exciton coupling, increases p-p distances and ultimately hinders hydrogen bonding of the carboxylic acids (Scheme 2). This hypothesis is in line with the formation of disordered aggregates through an isodesmic mechanism (Figure 3d), suggesting the absence of cooperative non-covalent interactions,that is,hydrogen bonding.
On the contrary,f or derivative 1,t he placement of the carboxyl group directly opposite of the TEG chains enables simultaneous pairwise interactions of the carboxylic groups via strong hydrogen bonding and 1D face-to-face stacking in the hydrophobic interior of the resulting nanostructure (see semiempirical calculations in Scheme 2a nd Figure S23). Therefore,t he lack of competition between entropic and enthalpic contributions in this system prevents pathway complexity.
To substantiate our packing model, the existence of hydrogen bonding in both 1 and B was probed by Fouriertransform infrared spectroscopy (FTIR) comparing thin films of the aggregate solutions in water and their respective monomer solutions in DCM. Prior to these measurements,we demonstrated by UV/Vis studies that the corresponding thermodynamic aggregates of 1 and B are retained upon thin-film formation ( Figure S24). Additionally,r eference compounds (4-bromobenzaldehyde and 4-iodobenzoic acid) were also measured in DCM to identify the position of the most relevant bands ( Figure 6). Compounds 1 and 2 in DCM ( Figure 6, middle panel) exhibit characteristic carbonyl stretching bands of conjugated aldehydes (ñ % 1710 cm À1 ) and carboxylic acids (ñ % 1735 cm À1 )t hat correspond to free, non-hydrogen-bonded groups.T he position of these bands coincides approximately with that of the reference compounds (upper panel in Figure 6), indicating the existence of non-hydrogen-bonded carbonyl groups in all investigated molecules.Only for 2 in DCM, an additional band is observed at % 1687 cm À1 (Figure 6, middle panel), which can be explained by the presence of hydrogen-bonded carboxylic groups resulting from the high concentrations (0.5 mm) required for the FTIR measurement.
In contrast, for thin films of aggregate 1,t he carbonyl stretch of the free,n on-H-bonded carboxylic acid groups disappears and as ingle band at ñ = 1689 cm À1 emerges ( Figure 6, bottom panel), which is ad istinctive feature of hydrogen-bonded closed dimers of carboxylic acids. [25] This behavior changes considerably for aggregate B,w here the carbonyl stretch of the carboxyl groups splits into two bands at ñ = 1721 cm À1 and 1694 cm À1 (open dimer Ia nd II, respectively;F igure 6, bottom panel). These results can be explained by an open acid-dimer interaction pattern in which only one of the carbonyl groups of the dimer acts as ahydrogen-bond acceptor in an interaction with the carboxyl OH of the second molecule,asdepicted in Figure 6, top.T he carbonyl group of the second molecule is consequently not involved in this interaction and therefore shifted to higher wavenumbers (open dimer I, ñ % 1721 cm À1 ). Thes light shift of this band compared with the carbonyl stretching frequency of the free acid (ñ % 1735 cm À1 )c an be explained by as mall electron-withdrawing effect on the C = Og roup due to the involvement of the carboxyl OH as ah ydrogen-bond donor. Such an open acid-dimer interaction has been previously identified for other systems in literature, [26] showing identical patterns as in the present aggregates of B.T hese results help to rationalize the flexible nature of the fibers of B compared to the much more rigid nanostructures formed by 1,s ince ac losed dimer leads to drastic restrictions in the degrees of freedom, while an open dimer is highly dynamic. [27] These findings support the previous hypothesis of an efficient hydrogen bonding arrangement for 1 that cannot be fully maintained for B without creating some defects,i n agreement with the appearance of two energetic states in the packing of B in emission studies ( Figure S12). Them ore favorable dimerization of 1 initially drives the self-assembly into amore planar, better pre-organized p-system, where the hydrophobic aromatic groups can be efficiently shielded from the aqueous environment upon extended aggregation. In contrast, due to geometrical constraints, 2 is unable to form sufficiently stable closed hydrogen-bonded carboxyl dimer seeds for extended supramolecular growth without exposing the hydrophobic content to the aqueous media. As ar esult, strong hydrophobic and aromatic interactions play am ore prominent role than hydrogen bonding in the thermodynamic assembly of B.T his is supported by the more significant broadening of the aromatic signals of 2 compared to 1 in 1 HNMR studies when D 2 Oi sa dded to [D 8 ]THF (Figure S25,S26). Forb oth 1 and B,t he aldehyde signal exhibits no deshielding,i ndicating the negligible role of this group in stabilizing both assemblies by hydrogen bonding,a sa lso supported by FTIR ( Figure S24).

Conclusion
In conclusion, we have reported an ew molecular design strategy for pathway control in aqueous self-assembly that relies on the modification of the relative position of the substituents by keeping the molecular geometry of the building blocks unaltered. To this end, we have investigated the aqueous self-assembly of two positional isomers of amphiphilic BODIPY dyes that solely differ in the position (2-or meso-) of two functional groups (carboxyl and aldehyde). An umber of experimental techniques (UV/Vis absorption and emission, FTIR, ITC,NMR, DLS,TEM, and cryo-TEM) complemented by theoretical calculations reveal that the relative position of the carboxyl groups with respect to the glycol side chains (either about 908 8 or 1808 8)governs the existence of pathway complexity:typical pairwise interaction of the carboxyl groups occurs efficiently (in al ocal hydrophobic microenvironment) only if the carboxyl group is placed in-line with the hydrophilic glycol side chains,resulting in as ingle thermodynamic cooperative lamellar assembly.In contrast, attachment of the carboxyl group at the mesoposition induces competition of entropic and enthalpic contributions,l eading to ak inetically controlled isodesmic pathway that rearranges over time into thermodynamically controlled nanofibers.F or this assembly,simultaneous J-type exciton coupling of the BODIPY dyes and partial hydrogen bonding of the carboxyl groups induce ac ooperative growth mechanism. Our results reveal that rational placement of functional groups for weak interactions is ap rerequisite for pathway selection in aqueous self-assembly.W eb elieve that our design strategy could be extended to various types of molecular platforms where substituents can be arranged in av ariety of defined angles.