Distinctive Synthetic Outcome of Non-Covalent Crosslinkers after Folding the Polymers

For the fabrication of DCLs that enable the synthesis of dynamic combinatorial non-covalent crosslinkers, aimed at directing the folding of polymeric templates, we initiated our approach by synthesizing a linear cationic homopolymer, poly[3-(methacryloylamino)propyl]trimethyl ammonium chloride (PT₇₀₃, Figure 1A), to serve as the folding scaffold. The inherent positive charge of the monomer predisposes the polymer at a stretched nascent conformation when in an aqueous solution, priming it for subsequent folding. The obtained polymer was characterized by proton nuclear magnetic resonance spectroscopy (¹H NMR, Figure S1) and size exclusion chromatography (Figure S2, Table S1). In the next step, we synthesized a benzoic dithiol building block, labelled A, featuring a carboxylic acid group (Figure 1A, Figure S3 and S4).¹¹ Upon deprotonation under physiological conditions, A acquires a negative charge and its thiol groups can be oxidized into disulfide bonds to yield macrocyclic species A_n, which function as potential non-covalent crosslinkers for polymer folding. The interaction between A_n and PT₇₀₃′s pendants, leading to charge neutralization, triggers the polymer‘s folding through hydrophobic effect.

Subsequently, two DCLs were made from A (3.00 mM) without and with the polymeric template PT₇₀₃ (15.00 mM, in monomer) in phosphate buffer at pH 7.4. In accordance with the concentration of A and PT₇₀₃, the two DCLs were named as 3A and 3A-15PT₇₀₃, respectively. The high-performance liquid chromatography (HPLC) spectra of the DCLs showed no noticeable changes two weeks after preparation. Even after one year, re-examination of the library revealed undetectable changes, indicating the chemical equilibrium of the A_n distribution remained stable. The HPLC-mass spectroscopy (HPLC-MS) analysis of the two equilibrated libraries suggested that the presence of PT₇₀₃ significantly supressed the production of A₄, while amplifying the production of the group of larger macrocycles ranging from the heptamer (A₇) to the hendecamer (A₁₁) (Figure 2A, Figure S5–S9). However, the HPLC-MS analysis of a control library prepared from A (3.00 mM) and a small molecular template ST (15.00 mM) (Figure 1A, Figure S10), an analogue to the pendant of PT₇₀₃, showed a similar spectrum as the library 3A (Figure S11). These results suggested that the polymeric template exhibited a unique template effect, which was not achievable with the monomeric template, even if it had the same recognition site.

This evident template effect encouraged us to verify whether PT₇₀₃ was bound with and folded by the A_n with various sizes using diffusion-ordered ¹H NMR spectroscopy (DOSY). In the DOSY two-dimensional (2D) plot of the fully oxidised 3A-15PT₇₀₃ library, the ¹H NMR signal from both A_n and PT₇₀₃ were broadened and in alignment at a same diffusion coefficient (D), indicating their quantitative supramolecular complexation (Figure 2B). The D of the supramolecular complex was larger than that of the 15.00 mM PT₇₀₃ (15PT₇₀₃) alone. By converting D to hydrodynamic radius (R_h) using Stokes–Einstein equation, a decreased R_h from ~6.86 nm of 15PT₇₀₃ to ~4.67 nm of 3A-15PT₇₀₃ was shown, suggesting folding of PT₇₀₃. Moreover, we determined the radius of gyration (R_g) of the polymer by Guinier analysis based on small angle X-ray scattering (SAXS) measurement, and compactness of the polymer by transverse NMR relaxation time (T₂) by ¹H NMR . In principle, smaller R_g represents smaller root mean square distance from the monomers to the polymer's centre of the mass. Shorter T₂ is an indicator of a more compact microenvironment around the molecule.¹² Indeed, PT₇₀₃ had significantly smaller R_g and shorter T₂ in the 3A-15PT₇₀₃ library than it alone (Figure 2C, Figure S12), confirming the polymer folding by A_n. In addition, D and T₂ of the library no longer changed observably after the chemical equilibrium, indicating a dual equilibrium between polymer folding and DCC. Lastly, the transmission electron microscopy (TEM) and dynamic light scattering (DLS) analysis showed that the supramolecular single chain nanoparticles (Supra-SCNPs) in 3A-15PT₇₀₃ library were compact globules with an average radius of 7.5 nm (Figure 2D and inset), which was consistent with that determined by R_h and R_g.

All together, these results confirmed that PT₇₀₃ was successfully folded into Supra-SCNPs by A_n synthesized from the 3A-15PT₇₀₃ library. The unique template effect in the DCLs suggested a possible structural adaptation of A_n to the chemical environments emerged from the folding.

Chemical Environments around the Non-Covalent Crosslinkers in the Supra-SCNPs

Given that polymer folding is propelled by hydrophobic compaction — a process that potentially diminishes the dynamism of involved motifs — we sought to identify these environments through the assessment of molecular dynamics (expressed as D and T₂) via ¹H NMR . Unfortunately, the ¹H NMR signals for A_n species initially coalesced, but separation was achieved by incorporating 30 % CD₃CN by volume in D₂O. This adjustment enabled precise assignment of the H₁ and H₂ signals for A₃, A₄, and the coalesced A₅-A₁₁ species, corroborated by DOSY (Figure 3A, Figure S13A), ¹H-¹H TOCSY (Figure S13B), and cross-referenced with HPLC data (Figure S13C). It is worth noting that the addition of CD₃CN preserved the supramolecular interaction within the Supra-SCNPs, as evidenced by the comparative T₂ values of A_n and PT₇₀₃ (Figure 2C), and Nuclear Overhauser Effect (NOE) correlations between A_n and PT₇₀₃ in the ¹H-¹H NOESY 2D plot (Figure 3B).

The discrimination of the NMR spectral peaks corresponding to different non-covalent crosslinkers facilitated the elucidation of their respective chemical environments via molecular dynamics. DOSY plots revealed that species A₅-A₁₁ aligned with PT₇₀₃, in contrast to A₃ and A₄ which exhibited greater diffusion coefficients. This gradation in T₂ values from A₃ and A₄ to A₅-A₁₁ (Figure 2C) intimates a deceleration in diffusion, indicative of three discrete chemical environments distinguished by their increasing degrees of spatial compactness. Crucially, A₃ and A₄ demonstrated marked NOE interactions with water protons (Figure S14), suggesting their localization within the hydrated shell of the Supra-SCNP. Thus, our findings categorize two primary chemical environments within the Supra-SCNPs, delineated by the compactness and specific binding locales: A₃ and A₄ were associated with more open binding sites within the hydrated shell, whereas A₅-A₁₁ were localized to the compacted binding sites within the hydrophobic core (Figure 3C).

The Adaptation Mechanism of Non-Covalent Crosslinkers in the Identified Chemical Environments

Incorporating this identified spatial distribution and the concentration distribution of library species, we hypothesized that the adaptive synthesis of A_n molecules was influenced by the increasing spatial compactness from the shell to the core of the Supra-SCNP. This gradation in compactness led to an augmented local effective concentration of A_n due to volumetric compaction. Guided by Le Chatelier's principle,¹³ which posits that an equilibrium system will adjust to minimize concentration changes, we anticipated a reduction in the number of core macrocycles through enhanced conversion into larger macrocycle variants.

To validate the implications of Le Chatelier's principle, we initially established a series of control DCLs without any template, utilizing variable concentration of A, to observe its influence on A_n’s distribution via HPLC analysis (Figure 3D). With increasing A’s concentration, a pronounced trend favouring the formation of larger macrocycles (A₇-A₁₁) over smaller ones (A₃ and A₄) was noted.

Further investigation assessed whether this principle could elucidate the observed preferential amplification of larger macrocycles in the context of the folding polymer PT₇₀₃. By modulating the degree of polymer folding to enhance Supra-SCNP volumetric compaction, we aimed to ascertain whether this would facilitate a higher conversion to larger macrocycles. Accordingly, a new series of libraries were prepared with a fixed concentration of PT₇₀₃ (15.00 mM, in monomer) and varying concentration of A to generate Supra-SCNPs at distinct folding intensities. As A’s concentration escalated from 0.30 mM to 7.50 mM, both the hydrodynamic radius (R_h) and the transverse relaxation time (T₂) of the Supra-SCNP diminished, indicative of a heightened folding degree and compactness (Figure 3E). Consistent with expectations, HPLC analysis revealed a proportional increase in the conversion to larger macrocycles correlating with the folding degree (Figure 3F). In comparison to control DCLs, the presence of the polymer template significantly augmented the conversion to larger macrocycles at a comparatively lower increment of A’s concentration, underscoring the folding-induced enhancement in local effective concentration of A_n. This finding corroborates the notion that macrocyclic non-covalent crosslinkers dynamically adapt to their folding milieu by counterbalancing the augmented local effective concentration resulting from folding, aligning with Le Chatelier's principle.¹⁴

To verify the significance of this structural adaptability from the crosslinkers, PT₇₀₃ was interacted with small molecules analogous to A_n. We chose benzoic acid (BA) and perylene-3,4,9,10-tetracarboxylic acid (PTA) with the same crosslinking motif of BA as A_n, but with elementary motif valences and immutable chemical structures bearing no structural adaptability. Compared to molecular behaviours induced by the interaction between A_n and PT₇₀₃, neither BA nor PTA folded PT₇₀₃, which was confirmed by the DLS analysis, DOSY 2D plots and T₂ measurements (Figure S15 and Figure S17, Table. S3). Moreover, according to the DOSY 2D plots, NOESY and T₂ measurements (Figure S17 and S18, Table. S3), the binding between the analogous crosslinkers and PT₇₀₃ was undetectable or weak. The more detailed data analysis has been shown in the support information (Materials and Methods, Figure S15–S18 and Table. S3). These results demonstrated the significance of the structural adaptability in the crosslinkers to drive efficient binding and folding on the polymer.

Interactive synchronisation of Self-Adaptive Synthesis and Dynamic Folding

Upon clarifying the spatial distribution and amplification mechanism of non-covalent crosslinkers within Supra-SCNPs at equilibrium, our investigation proceeded to the synchronized dynamics of self-adaptive synthesis and polymer folding.

First, we need to profile both the DCC and folding kinetics synchronously in a library until the dual equilibrium. To achieve this, we monitored the kinetics of compositional changes of A_n in the 3A-15PT₇₀₃ library using HPLC. Simultaneously, the folding kinetics of the polymer template was tracked by DOSY (Figure 4A, Figure S19).

At the beginning when the library was prepared, A already bound to PT₇₀₃ quantitatively, evidenced by their common D determined by DOSY (Figure S20), suggesting the interactive dynamics between two processes from the start. As the oxidation progressed, A₃ and A₄ emerged as the primary and secondary products, respectively, up to 200 hours after preparing the library. In parallel, the R_h of PT₇₀₃ decreased and reached a plateau, indicating a kinetic intermediate state. At this stage, as the thiol almost entirely consumed, the disulfide-disulfide exchange reaction initiated the conversion from A₃ and A₄ to A₅-A₁₁. As the exchange reaction approaching equilibrium, the concentration of larger macrocycles A₅-A₁₁ significantly increased, dominating the production of crosslinkers. Concurrently, the size of polymer PT₇₀₃ reduced and stabilized accordingly. These kinetic studies demonstrated two stages of synchronization between reversible chemical and folding process.

The first stage of synchronization was marked with the presence of thiol and dominated by the thiol-disulfide exchange. We hypothesised that the thiols inhibited the formation of A₅-A₁₁. If the hypothesis is correct, introducing fresh A to a pre-equilibrated library will transiently reverse the conversion of A₅-A₁₁, which will then gradually reform upon the oxidation of thiols. To verify, 0.75 mM fresh A was introduced to a pre-equilibrated 3A-15PT₇₀₃ library (Figure 4B). Indeed, the larger macrocycles A₅-A₁₁ were almost quantitatively converted to A₃ right after the addition. Then the gradual conversion from A₃ to A₅-A₁₁ followed the oxidation of free thiols. Moreover, we testified the instant dual equilibrium after mixing pure A₃ or A₄ (3.00 mM, in A) with 15.00 mM PT₇₀₃ by examining the chemical equilibrium by HPLC and folding equilibrium by DOSY (Figure S21). The libraries prepared by these two mixing procedures instantly reached the same A_n distribution and D as the reference library equilibrated from mixing fresh A and PT₇₀₃, and no longer changed, indicating reaching the same dual equilibrium. It validated the premise that A₃ and A₄ binding to the polymer inherently promoted the dual equilibrium by disulfide-disulfide exchange, yet free thiols prevented the synthesis of A₅-A₁₁ and folding progression by thiol-disulfide exchange.

Notably, the second phase‘s kinetic profile suggested a catalytic synchronization between DCC and folding. It was primarily indicated by the sigmoidal transition from A₃ and A₄ to A₅-A₁₁. Since A₃ and A₄ were probed with stronger dynamics and hydrophilicity than A₅-A₁₁ when binding with PT₇₀₃ (Figure 2C and S14), we reasoned that the binding of A₃ and A₄ would not only introduce hydrophobic effect by charge neutralisation, but also catalyse the chain rearrangement from the initial aqueous environment, making it kinetically accessible to the folding that formed the hydrophobic core. In turn, the hydrophobic core amplified A₅-A₁₁ at the consumption of A₃ and A₄ governed by the Le Chatelier's principle, marked as a local energy minimum for synchronisation. As a result, A₃ and A₄, when being synchronised with template folding, worked as both reactants and catalysts for their own reaction to A₅-A₁₁ (Figure 4 C). Along with the continued folding and consumption of A₃ and A₄ during this synchronisation, the concentration of reactant and catalytic activity both dropped, which resulted the sigmoidal kinetic profile toward equilibrium.

Additionally, though sharing the similar sigmoidal synthetic profile, this catalytic synchronisation toward thermodynamic equilibrium differed from the self-catalytic replicators in out-of-equilibrium systems.¹⁵ Self-replications are more akin to biological reproduction. Catalytic synchronisation between DCC and folding, through its critical dependence on the timing of product formation, more closely mirroring biotic peptide folding introduced earlier. The criticality of timing was probed by isolating a mixture of A₅-A₁₁ from a balanced 3A-15PT₇₀₃ library via preparative HPLC, subsequently introducing it to PT₇₀₃ (15.00 mM). Contrary to Supra-SCNP formation, precipitation observed (Figure 4D), presumably due to robust interchain interactions facilitated by A₅-A₁₁′s affinity for the polymer. Coupled with previous NOESY, DOSY and T₂ analysis indicating A₃ and A₄ being on the outer shell of the Supra-SCNPs, we hypothesised that the bound A₃ and A₄ on polymer also worked as surfactants for the hydrophobic core. Indeed, when we mixed A₃ and A₄ together with the larger ones A₅-A₁₁ and the polymer, the dual equilibrium was reached instantly (Figure S21).

These findings demonstrated the synchronized adaptive synthesis of non-covalent crosslinkers A_n and the folding process, which guaranteed the efficient thermodynamic folding control: A₃ and A₄ preorganised the polymer into an intermediate and single-chain state, catalysing the folding, and folding further facilitated the disulfide-disulfide exchange reaction from A₃ and A₄ to A₅-A₁₁, thus stabilising the folding and Supra-SCNP formation (Movie. S1). Without the pre-organization of A₃ and A₄, the stabilized binding and hydrophobic folding between A₅-A₁₁ and PT₇₀₃ would inevitably happen at an inter-chain level, thus leading to inter-chain aggregation instead of single-chain folding.

Control the Self-Adaptive Synthesis and Folding for Potential Applications

This understanding spurred our exploration into how the integration of external functional molecules could modulate the adaptive synthesis and enable the functional emergence of Supra-SCNPs. Given the hydrophobic-driven folding mechanism, we hypothesized hydrophobic anti-cancer drugs, such as doxorubicin (DOX), could be accommodated within Supra-SCNPs through an adaptive folding process, marking significant strides toward drug delivery systems (DDSs). Moreover, the swift attainment of thermodynamic equilibrium, in the absence of unreacted thiols, underscores a strategic approach for preventing drugs from denaturing during the fabrication process of DDSs.

Experimentally, DOX was homogenized with 15 mM PT₇₀₃ at a final concentration of 2.3 mM. The solution was then mixed with a pre-equilibrated DCL made by 3.00 mM A to reach the equilibrium instantly, ensuring the maximum drug loading content (DLC) of the Supra-SCNPs loaded with fresh DOX (Supra-SCNP-DOX). Using HPLC-MS, the DLC of the Supra-SCNP-DOX was determined as 22 % with a loading efficiency of 86 % and no side product was found than the macrocycles and DOX (Figure S22A). Interestingly, A₃ was more amplified in Supra-SCNP-DOX, which proposed an adaptive enlarging of folding after loading the cargo molecules since he low-degree folding favored A₃. Indeed, the TEM and DLS results suggested that the average R_h of the Supra-SCNP-DOX was increased to approximately 33 nm (Figure S22B and C). These results clearly suggest that we can use the external molecules to control the self-adaptive folding of the Supra-SCNP-DOX synthesized from DCLs.

Advancing to practical applications, we evaluated the potential of Supra-SCNP-DOX as a DDS. Acknowledging the elevated glutathione (GSH) levels and the lower intracellular pH (~5.6) characteristic of cancer cells,¹⁶ we posited that such conditions would facilitate the disassembly of Supra-SCNP-DOX by reducing the disulfide bonds and protonating the carboxylate groups, enabling the release of encapsulated drugs. The hypothesis was confirmed as DOX release from Supra-SCNP-DOX was significantly expedited under reductive or acidic conditions (Figure 5A), showcasing the system‘s responsive drug delivery capability.

Assessing the biosafety of Supra-SCNP-DOX, we conducted in vitro cytotoxicity tests against multi-drug-resistant ovarian cancer cells (NCI/RES-ADR). The Supra-SCNP-DOX displayed markedly superior anti-cancer efficacy compared to free DOX and the polymer-DOX mix at equivalent concentrations (Figure 5B). Interestingly, among all the dosing configurations, Supra-SCNPs exhibited minimal cytotoxicity, while polymer‘s inherent toxicity was even more significant than the polymer-DOX mix (Figure 5C), underscoring the safety of the carrier component, and emphasising the Supra-SCNP′s enhanced drug delivery efficacy against drug-resistant cancer. Flow cytometry and confocal laser scanning microscopy (LCSM) further validated the superior drug delivery efficiency of Supra-SCNP-DOX, indicating a substantial uptake of DOX by cancer cells (Figure 5D and E), thereby affirming the system‘s efficacy in overcoming drug resistance through optimized drug delivery.