Self-Assembling Ability Determines the Activity of Enzyme-Instructed Self-Assembly for Inhibiting Cancer Cells.

Enzyme-instructed self-assembly (EISA) represents a dynamic continuum of supramolecular nanostructures that selectively inhibits cancer cells via simultaneously targeting multiple hallmark capabilities of cancer, but how to design the small molecules for EISA from the vast molecular space remains an unanswered question. Here we show that the self-assembling ability of small molecules controls the anticancer activity of EISA. Examining the EISA precursor analogues consisting of an N-capped d-tetrapeptide, a phosphotyrosine residue, and a diester or a diamide group, we find that, regardless of the stereochemistry and the regiochemistry of their tetrapeptidic backbones, the anticancer activities of these precursors largely match their self-assembling abilities. Additional mechanistic studies confirm that the assemblies of the small peptide derivatives result in cell death, accompanying significant rearrangement of cytoskeletal proteins and plasma membranes. These results imply that the diester or diamide derivatives of the d-tetrapeptides self-assemble pericellularly, as well as intracellularly, to result in cell death. As the first case to correlate thermodynamic properties (e.g., self-assembling ability) of small molecules with the efficacy of a molecule process against cancer cells, this work provides an important insight for developing a molecular dynamic continuum for potential cancer therapy, as well as understanding the cytotoxicity of pathogenic assemblies.

Introduction

Only approximately 3000 of the predicted ∼30 000 genes in the human genome are coded for proteins that possess the ability of binding small druglike molecules.[1,2] Among the 3000 druggable genes, only 600–1500 are disease-associating genes that are potential drug targets.[1] The limited number of small-molecule drug targets urge the development of innovative approaches other than tight ligand–receptor binding.[3] As a complementary process for ligand–receptor interactions, enzyme-instructed self-assembly (EISA) is a ubiquitous phenomenon in cellular processes that affords spatiotemporal control of higher order structures from nanoscales to microscales.[4] Inspired by such a fundamental fact in cell biology, we and others are employing EISA of small molecules to develop new therapeutics,[5−7] especially for cancer therapy.[8−11] Generating supramolecular assemblies via EISA enables selective targeting of the undruggable targets or simultaneous interaction with multiple targets.[12] For example, alkaline phosphatase (ALP), being reported as a biomarker of cancer for about 5 decades,[13] remains undruggable due to the difficulties in achieving inhibitor selectivity and sufficient cell permeability.[14] Recently, we have selectively targeted such a undruggable feature on cancer cells via EISA of small peptides.[15] Moreover, the supramolecular assemblies formed via EISA not only inhibit cells via multiple mechanisms but also promise to prevent acquired drug resistance.[16] In addition, EISA provides an effective approach for targeting loss-of-function (i.e., silencing tumor suppressors) in cancer cells,[17] which ultimately may meet such a major challenge in translational medicine. Several other laboratories also pioneered the exploration of EISA for biomedical applications, including inhibiting cancer cells. For example, Pires et al. demonstrated using EISA of a carbohydrate derivative to selectively inhibit osteosarcoma cells that overexpress ALP.[18] Maruyame and co-workers employed a protease (e.g., MMP-7) to trigger molecular self-assembly of a peptide lipid and to induce cancer cell death through intracellular EISA.[19] Moreover, Liang and co-workers, combining EISA with GSH-controlled condensation, used one precursor to differentiate the extra- and intracellular environments to yield two different nanofibers via self-assembly.[20] Yang and colleagues demonstrated the use of enzyme-catalyzed hydrogel as an efficient adjuvant to boost immune response to a vaccine.[21] In addition, EISA also finds applications for photoacoustic imaging of furin-like activity[22] and monitoring autophagy.[23] Moreover, the concept of EISA is applicable to nanoparticles.[24] These studies not only expand the scope of precursors and enzymes utilized for EISA but also underscore the promises of EISA for a variety of biomedical applications.

Despite the promise of EISA in selectively inhibiting cancer cells or molecular imaging,[25] there are several important questions that remain to be answered. Of special significance is how to design a small molecule for EISA. That is, what molecular feature (or thermodynamic property) is the most important factor for increasing the efficacy of EISA for various applications? Particularly in the case of small peptides, although it is conceivable to obtain the different activities of peptide assemblies by varying the residues, sequences, or capping group of peptides or peptide derivatives,[26] the relationship between the self-assembling ability of small molecules and the corresponding activity remains to be established, which is crucial for guiding the design of small molecule assemblies for anticancer therapy, as well as for other applications. To address this critical question, we designed and synthesized a series of structural analogues of peptidic precursors (Scheme ) that differ in several key molecular features: C-terminal capping, stereochemistry, and regiochemistry. Our results indicate that the self-assembling abilities of these peptide derivatives dictate the anticancer activity of EISA for inhibiting cancer cells. That is, the precursors with higher self-assembling ability turn into the molecules (e.g., hydrogelators) that exhibit higher self-assembling ability, resulting in effective formation of assemblies that inhibit cancer cells. Cell imaging reveals that the assemblies of the small peptide derivatives likely affect the interactions between plasma membrane and cytoskeletons to result in cell death. Besides positively corroborating the inhibitory activity and self-assembling ability of small molecules, this work also provides insights for understanding in situ molecular self-assembly in cell milieu[27] and helps develop EISA as a molecular process for potential cancer therapy, as well as other biomedical applications.

Scheme 1

Molecular Structures of the Precursors and the Correlation between the Ability for Self-Assembly of Small Molecules and Anticancer Activity

Results and Discussion

Molecule Design and Synthesis

On the basis of an anticancer precursor (1p)[28] of EISA and via fragment combination and the mutation of peptide sequences, we design a series of EISA precursors that consist of a peptidic backbone, an ALP cleavage site, a carboxylic modification, and an N-terminal capping (Scheme ). The peptidic backbone d-Phe-d-Phe is the enantiomer of the well-studied dipeptide Phe-Phe, which forms nanocrystals[29] mainly stabilized by aromatic–aromatic interactions. The phosphotyrosine residue provides enzymatic cleavage site for ALP to generate supramolecular assemblies.[30] Because O-methylation of protein increases the hydrophobicity and neutralizes the negative charge of amino acid residues, which results in enhanced self-assembling ability[11] and nonspecific binding to cell membrane,[31] we choose to O-methylate the carboxylic acid groups in d-Glu, l-Glu, and d-Asp. The difference in stereochemistry of C-terminal amino acid residue (i.e., d-Glu and l-Glu) or in side-chain length (i.e., d-Glu and d-Asp) would verify whether conventional ligand–receptor binding contributes to the activities of the assemblies. Considering that the carboxyl ester is a substrate of esterase and a receptor of nucleophilic attack, we use N-methylacetamine (−CONHMe) as the C-terminal capping of the peptide. Thus, the difference between −COOMe and −CONHMe groups would delineate the contribution of the reactivity of the assemblies for the activities. Inspired by protein N-acetylation, a phenomenon that occurs in almost all eukaryotic proteins,[32] we cap the N-terminal of the precursors by a 2-naphthylacetyl group, since naphthyl groups provide strong intermolecular aromatic–aromatic interactions.

By systemically combining the aforementioned peptide backbone, N-terminal capping, and C-terminal modifications, we would generate five new precursors (Scheme , 2p, 3p, 4p, 5p, and 6p). These precursors differ from 1p(28) in terms of N-terminal capping, C-terminal modification, stereochemistry, or regiochemistry, which would help answer the following questions: (1) How do the structural differences (i.e., N-terminal capping, C-terminal modification, stereochemistry, and regiochemistry) affect the nanostructures of assemblies via EISA? (2) How do the self-assembling abilities of either the phosphorylated precursors or the dephosphorylated products determine the efficacy of EISA for inhibiting cancer cells?

The designed precursors and their corresponding self-assembling molecules are accessible via a facile synthetic route [Scheme S1, Supporting Information (SI)]. We first prepared phospho-d-tyrosine in 90% yield,[33] followed by the N-Fmoc protection, which yielded Fmoc-Tyr(PO3H2)-OH.[34] Following a general procedure for solid-phase peptide synthesis,[35] we synthesized N-terminal capped peptides with Fmoc-protected amino acids. Esterification of the carboxylic acid groups, catalyzed by trimethylsilyl bromide,[36] produced the O-methylated precursors. We directly coupled the carboxylic acids with methylamine to obtain the N-methylamide derivative (5p). NMR spectra and LC–MS confirmed the structures of precursors after high-performance liquid chromatography (HPLC) purified them.

Self-Assembling Ability in Vitro

To evaluate the self-assembly of the precursors and the corresponding self-assembling molecules in vitro, we employed transmission electron microscopy (TEM) to visualize the nanostructures formed by the precursors, at 0.5 wt %, before and after EISA occurs. As shown in Figure , dissolving in aqueous solution and bearing d-glutamic acid diester at its C-terminal, 1p mainly forms short nanofibers with a width of 6 ± 2 nm and a length of 60 ± 10 nm, together with several thicker nanofibers with a diameter of 12 ± 2 nm. Containing l-glutamic acid diester and as a diastereomer of 1p, 2p results in nanofibers with a diameter the same as that of 1p (i.e., 6 ± 2 nm), but the lengths of the nanofibers range from nanometers to micrometers. The replacement of d-glutamic acid diester with d-aspartic acid diester yields 3p, which shows different self-assembly morphology, containing mainly longer fibers. These fibers are typically micrometers long and have a diameter of 6 ± 2 nm. The difference between 1p and 3p likely arises from the length of the side chain of the amino acid. Precursor 4p, inserting the d-glutamic acid ester between the phenylalanine and tyrosine residues and having an O-methylated C-terminal, results in more uniform long fibers several micrometers in length and with widths of about 7 ± 2 nm. Using methyl amine as the C-terminal modification group produces 5p, which forms nanofibers (6 ± 2 nm) interacting with amorphous aggregates. Using an acetyl group to replace a naphthyl motif in 1p results in 6p, which forms some big aggregates, together with thinner fibers (5 ± 2 nm in diameter) than those of 1p–5p. The naphthyl-capped precursor (e.g., 1p) has higher self-assembling ability than the acetyl-capped precursor (e.g., 6p), verifying that naphthyl groups provide strong aromatic–aromatic interactions. These results indicate that the precursors self-assemble to form certain nanostructure when being dissolved in aqueous solution at a physiological pH and at relatively high concentration (0.5 wt %).

Figure 1

TEM images of the nanostructures formed by 0.5 wt % 1p–6p in pH 7.4 water, before and after adding ALP (2 U/mL) (scale bar, 100 nm).

Upon the treatment with ALP, 1p turns into 1, which self-assembles to yield rigid nanofibers with a diameter of 14 ± 2 nm, similar as the previous observation.[28] After forming by dephosphorylation, 2, 3, or 4 self-assembles to form nanofibers with a diameter of 6 ± 2 nm, nanoribbons, and nanofibers with a diameter of 7 ± 2 nm, respectively, implying that stereochemistry, the side chain length, and regiochemistry all affect the morphologies of the self-assembled nanostructures generated by EISA. Unlike the case of 1p, the EISA of 5p generates long flexible nanofibers with diameters of 6 ± 2 nm, indicating that N-methylacetamide is less hydrophobic than the methyl ester at the C-terminal of the peptide and provides a slightly weaker self-assembling ability. Dephosphorylation of 6p generates 6, which forms straight nanofibers with a diameter of 9 ± 2 nm. Table summarize the morphologies of the assemblies without or with the addition of ALP.

Table 1

Summary of the Self-Assembly of the EISA Molecules

compounda	1p	2p	3p	4p	5p	6p
morphology before adding ALP (db, nm)	fibers (6 ± 2, 12 ± 2)	fibers (6 ± 2)	fibers (6 ± 2)	fibers (7 ± 2)	aggregate, fibers (6 ± 2)	aggregate, fibers (5 ± 2)
morphology after adding ALP (db, nm)	fibers (14 ± 2)	fibers (6 ± 2)	nanoribbons	fibers (7 ± 2)	fibers (6 ± 2)	fibers (9 ± 2)

The concentration is 0.5 wt %.

Diameter of nanofibers.

To more precisely evaluate the self-assembling ability in the context of EISA, we measured the critical micelle concentration (cmc) of the above precursors (2p–6p) and their corresponding self-assembling molecules (2–6) by using rhodamine 6G as a probe.[37] [Figures and S16 and S17 (SI)] Compared with other precursors, 1p bears the highest self-assembling ability (i.e., the lowest cmc of 30 μM[28]). The cmc values of 2p–4p are higher than that of 1p, indicating that the changes in the stereochemistry, side chain length, and regiochemistry all affect the self-assembling ability of the precursors. The cmc of 5p is more than double of that of 1p, indicating that the O-methylation of the tetrapeptide results in higher self-assembling ability than the N-methylamidation does. 6p exhibits nearly 70-fold weaker self-assembling ability, which is consistent with the TEM results above. The cmc values of the precursors follow the order of 1p < 3p < 2p < 4p < 5p < 6p. The self-assembling abilities of the dephosphorylated molecules follow the trend of 1 > 3 > 2 > 4 > 5 > 6, the same as that of the precursors. This result likely originates from the identical difference (i.e., phosphorylation) between the precursors and the self-assembling molecules. The TEM images reveal that there are hardly any nanostructures formed by the precursors at the concentration lower than the cmc of the corresponding self-assembling molecules, without or with the addition of ALP (Figure S18, SI). The correlation between the cmc and the fiber formation likely resembles the formation of worm micelles from micelles.[38] In fact, the size and morphology of the assemblies of 1 depend on the concentration of 1 (Figure S19, SI), which supports this inference.

Figure 2

Cmc values of precursors 1p–6p and their corresponding dephosphorylated peptide derivatives 1–6.

Anticancer Activity and Static Light Scattering

The assemblies formed via EISA are cytotoxic to cancer cells, while the monomers are innocuous,[7] suggesting the importance of the in situ self-assembling process for inhibiting cancer cells. These results prompt us to examine the correlation between the self-assembling ability of EISA molecules (i.e., the precursors and the self-assembling molecules) and their efficacy for inhibiting cancer cells. We choose an osteosarcoma cell line (Saos-2) as the cancer cells for the test because the high expression level of ALP on Saos-2 warrants fast dephosphorylation of the precursors.[18,39] The thermodynamic parameter (−ΔG0) characterizes the free energy change for formation of assemblies,[40] and the pIC50 values, which is −log10(IC50),[41] represent the cytotoxicity of the precursors. As shown in Figure , the potency (pIC50) of EISA precursors at 24 h against Saos-2 cells follows the order of 1p > 2p > 3p > 4p > 5p > 6p, exhibiting positive correlation with the self-assembling ability (−ΔG0) of the precursors (and the dephosphorylation products) except the relative order of 2p and 3p, which, nevertheless, are quite close in both potency and self-assembling ability. At the concentration of their IC50 values, the precursors (1p–6p) hardly form any nanostructures in aqueous solution. However, the addition of ALP into the above solutions results in irregular fibrous structures, indicating the formation of assemblies via EISA at these concentrations (Figure S21, SI). While all the designed molecules assemble into fibrous structures, these nanostructures differ slightly in morphology. We speculate that the differences between the morphologies of the nanostructures may marginally contribute to the subtle cytotoxicity differences between the EISA molecules. In fact, the IC50 values of the precursors correlate with the cmc values of corresponding self-assembling molecules, indicating that the in situ formation of assemblies plays a critical role in inhibiting cancer cells. These results suggest that the self-assembling ability of EISA precursors and their self-assembling products determine the potency of EISA against cancer cells.

Figure 3

Correlation between the self-assembling ability (−ΔG0) and anticancer activity (pIC50) of EISA molecules against Saos-2 cells.

The positive correlation between the self-assembling ability of the molecules for EISA and their anticancer efficacy suggests that the amounts of assemblies generated in situ (i.e., on or inside cancer cells) may be critical in inhibiting the cancer cells. To test this hypothesis, we measured the static light scattering (SLS) of the solutions of three representative precursors (i.e., 1p,[28]5p, and 7p(28)) (Scheme S2, SI) before and after the addition of ALP, since the signal intensity is proportional to the amount of assembly. We choose to compare these precursors because they share the same backbone structure and only differ in the modification of the carboxylic group, which is methyl ester for 1p, methyl amide for 5p, and carboxylic acid for 7p. As shown in Figure A, the light-scattering results show that the increase of the SLS signal depends on the concentrations of the precursors. The SLS signal of 1p increases from 0.04 to 1.80 as the concentration rises from 5 to 50 μM, indicating that more assemblies formed at higher concentrations. The solutions of 5p and 7p exhibit little SLS signal, even at the concentration of 50 μM, suggesting that the precursors hardly form any assemblies at these concentrations, which is consistent with their cmc results [Figures and S16 and 17 (SI)]. The addition of ALP to the solution of 1p and 5p causes significant increase of the SLS signals, confirming the generation of assemblies via EISA. However, the ALP treatment shows little effect on the solution of 7p, likely due to the weak self-assembling abilities of 7p and 7. After ALP treatment, the signal intensity of the solutions of the precursors follows the order of 1p > 5p > 7p, confirming that the amounts of assemblies of 1, 5, and 7, formed via EISA at the initial concentration of 5 μM or 50 μM, follow the trend of 1 > 5 > 7.

Figure 4

(A) Intensity of static light scattering (SLS) of the solutions of 1p, 5p, and 7p (5–50 μM) before and after adding ALP (1 U/mL) for 12 h in pH 7.4 PBS buffer (light-scattering angle = 60°). (B) The cell viability of Saos-2 cells treated with 1p, 5p, or 7p (5–50 μM) for 24 h.

To correlate the concentration of the assemblies with inhibition efficacy, we also tested the cell viabilities of Saos-2 cells treated with 1p, 5p, or 7p within these two concentrations, 5 and 50 μM (Figure B). The cell viability results show that 1p and 5p exhibit cytotoxicity in a dose-dependent manner, while 7p is innocuous to Saos-2 cells. At the same concentration, the cytotoxicities of the precursors follow the order of 1p > 5p > 7p at all concentrations, agreeing with the amount of assemblies formed via EISA (vide supra). When the concentration of 5p increases from 20 to 50 μM, there is a significant decrease of cell viability. This observation correlates well with the fact that there are significantly more assemblies at 50 μM 5p after dephosphorylation (Figure S22, SI). The increase of concentration results in the enhanced cytotoxicity of 1p and 5p, positively correlating with the increase of the assemblies. These results indicate that the self-assembling abilities of these molecules determine the efficacy of EISA for inhibiting Saos-2 cells because higher self-assembling ability results in more assemblies at a constant concentration of the precursors. Remarkably, 8.8 μg/mL of 1p inhibits over 90% of Saos-2 cells (Figure S23, SI). This IC90 value of 1p against Saos-2 cells is comparable to or, in fact, even more potent than the clinically used drug (cisplatin) for treating osteosarcoma.[42]

Inhibitory Activities against Different Cell Lines

In addition to Saos-2 cells, we also examined the cytotoxicity of 1p and 5p on different cancer cell lines [breast adenocarcinoma cells (MCF-7) and glioblastoma cells (T98G)] and a normal stromal cell line (HS-5). (Figure ) In contrast to the case of Saos-2 cells, 5p shows a higher inhibition effect than 1p on MCF-7 cells. The IC50 of 5p on MCF-7 is 47.0 μg/mL (53.5 μM), while the IC50 of 1p is 92.4 μg/mL (105 μM). This result likely results from the hydrolysis of the methyl ester bond of 1p, since MCF-7 cells express relatively high level of carboxylesterases.[9] In addition, the IC50 of 1p (>176 μg/mL) (200 μM)[28] on HepG2 cells, which are known to greatly overexpress esterase, is 2-fold the IC50 of 5p (92.4 μg/mL) (111 μM) (Figure S24, SI). The hydrolysis of the ester on 1 would generate 1″ (vide infra), which possesses much lower self-assembling ability. While the precursors potently inhibit Saos-2 cells with an IC50 value of 3.9 μg/mL (4.4 μM) for 1p and 38.9 μg/mL (44.3 μM) for 5p, they scarcely exhibit cytotoxicities to T98G or HS-5 cells, even at the concentration of 176 μg/mL (200 μM). This result agrees with the low expression level of ALP on these two cell lines,[10] further confirming that the selective inhibition of EISA precursors against cancer cells originates from the expression level of enzymes and the rate of self-assembly resulted from the enzymatic reaction.

Figure 5

IC50 (72 h) of 1p or 5p against Saos-2 cells, MCF-7 cells, T98G cells, or HS-5 cells.

Molecular Transformation in Cellular Milieu

To reveal the molecular transformation and self-assembly of EISA molecules in cellular milieu, we coincubated 1p or 5p with Saos-2 cells, HepG2 cells, or HS-5 cells and quantified the conversion of the precursors after 24 h incubation (Figure ). LC–MS and HPLC analyses show that the endogenous ALP turns 1p into 1, carboxylesterase (CES) converts 1p to 1p′, and catalysis by ALP and CES yields 1″, while the 5p only transforms into 5 by ALP, due to the stability of methyl amide (Figures S25–S27, SI). As shown in Figure A, 15.2%, 46.9%, and 62.2% of 1p molecules remain in Saos-2, HepG2, and HS-5 cells, respectively, indicating that the activities of ALP on these cells follow the order of Saos-2 > HepG2 > HS-5,[10] which is also consistent with the cytotoxicity of 1p for these cell lines. Although HepG2 exhibits higher ALP activity than HS-5, its high expression level of CES results in hydrolyzing 26% of carboxyl methyl ester, which is over twice of the same hydrolysis on HS-5 cells (i.e., 9.7%). As a result, the amount of self-assembling molecule 1 in the cultures for Saos-2, HepG2, and HS-5 are 75.5%, 27.1%, and 28.1%, respectively. Because of the much poorer self-assembling ability of 1″ compared with that of 1, these results explain that 1p potently inhibits Saos-2 cells, but it is innocuous to HepG2 cells.

Figure 6

(A) Percentage of compounds 1p′ (hydrolysis product of 1p), 1p, 1″, and 1 after incubating 1p (200 μM) with Saos-2 cells, HepG2 cells, or HS-5 cells. (B) Percentage of compounds 5p′ (hydrolysis product of 5p), 5p, 5″, and 5 after incubating 5p (200 μM) with Saos-2 cells, HepG2 cells, or HS-5 cells. The ratios were determined using HPLC and LC–MS. Cells were treated for 24 h.

In addition, ALPs on Saos-2, HepG2, and HS-5 cells convert 89.6%, 47.6%, and 30.0% of 5p to 5, respectively, further confirming that the activities of ALP follow the order of Saos-2 > HepG2 > HS-5. For 1p and 5p, Saos-2 cells dephosphorylate comparable amounts of the precursors (i.e., 66.2% for 1 and 75.5% for 5), supporting that the inhibitory efficacy follows the trend of self-assembling ability. For HepG2 cells, the total amount (i.e., 47.3%) of 1 and 1″ is almost same as the generation of 5 (i.e., 47.6%). However, the 1/1″ molar ratios in the cultures of HepG2 cells are 27.1/20.2. Together with the cytotoxicity data, these results suggest that the amount of the assemblies (made of 1 and 1″, or 5), in fact, determines the efficacy of EISA. These detailed results of the conversion of the EISA molecules, indeed, reveal the complexity of EISA in cell milieu, which underscores the importance of correlating the self-assembling ability of stable molecules with the efficacy of EISA for inhibiting cancer cells.

Mechanism of Cell Death

To confirm that the EISA processes play critical roles in the inhibitory effect, we coincubated the precursor 1p and exogenous ALP or a tissue-nonspecific alkaline phosphatase (TNAP) inhibitor[43] with Saos-2 cells. Both exogenous ALP and the TNAP inhibitor rescue the cells, increasing the cell viability from 11% to 69% and 23%, respectively (Figure S28, SI). This result validates the contribution of EISA processes. The TNAP inhibitor 2,5-dimethoxy-N-(quinolin-3-yl)benzenesulfonamide (DQB) only inhibits TNAP effectively,[43] so it is unlikely that it would completely rescue the cells due to the presence of other isozymes of ALP[44] on the cells or other phosphatases in cell milieu. The addition of pan-caspase inhibitor (zVAD-fmk)[45] or necroptosis inhibitor (Nec-1)[46] increases the cell viability of Saos-2 cells treated with the 1p to 25% or 32%, respectively (Figure S28, SI), indicating that either apoptosis or necroptosis represents a path to the death of Saos-2 treated by 1p, though other death pathways remain to be validated. In addition, Congo red staining results reveal that the precursors (1p and 5p) form assemblies both pericellularly and intracellularly via EISA (Figure S29, SI). Congo red is unable to stain the apoptotic Saos-2 cells induced by cisplatin (Figure S30, SI), further confirming that the staining of Saos-2 cells incubated with EISA molecules by Congo red originates from the assemblies of the EISA molecules (e.g., 1 or 5).

The EISA of 1p/1 also leads to the morphology change of Saos-2 cells, indicating that the resulting assemblies affect the integrity of the cytoskeletons. We stained the F-actin with Alexa Fluor 633 Phalloidin to reveal the shape changes. The staining results (Figure ) show that the treatments of precursors disrupt the F-actin arrangement, while the control Saos-2 cells exhibit stretched thin actin filaments. The treatment of 1p and 2p results in significant shrinkage of actin networks and much higher density of F-actin at the cell boundary. Notably, the actin filaments are much shorter in the cells treated with 2p. Similarly, 3p–5p also cause the accumulation of actin near plasma membranes, reflecting the increase of focal adhesion and the decrease of fibrillar adhesion of cells.[47] Different from the control cells, the actin filaments become much shorter or punctate upon the treatment with 3p, 4p, or 5p. However, 6p shows little effect on the actin of Saos-2 cells, likely due to the poor self-assembling ability of 6. The gradual change of the actin morphology upon the treatment of the precursors from 1p to 6p indicates that the assemblies of the peptide derivatives likely disrupt actin dynamics and cause the cell death.

Figure 7

CLSM images of Saos-2 cells stained with Alexa Fluor 633 Phalloidin (F-actin, red) and Hoechst (nuclei, blue) after the treatment of culture medium, 1p, 2p, 3p, 4p, 5p, or 6p for 12 h. Scale bars = 20 μm.

To get more insights on how the EISA process affects the cytoskeletons, we also evaluated the changes of microtubules upon the treatment of the precursors. As shown in Figure , 1p leads to the reorganization of microtubules in the proximity of plasma membranes of Saos-2 cells, which likely corresponds with the apoptotic microtubule network (AMN)[48] formed during the execution phase of apoptosis. This result supports that the treatment of 1p is able to cause apoptosis. The microtubules of Saos-2 cells become shorter and accumulate near the cell boundary after incubation with 2p or 3p. In addition, 2p, 3p, 4p, or 5p causes tubulins to aggregate at the centromere in Saos-2 cells, while 6p hardly influences the morphology of microtubules. Moreover, the live cell imaging shows that the addition of 1p rapidly leads to the membrane blebbing of Saos-2 cells and reorganization of plasma membrane followed by the loss of focal adhesion,[49] agreeing with the disruption of cytoskeleton dynamics resulting in cell death.

Figure 8

CLSM images of Saos-2 cells stained with tubulin tracker (green) and Hoechst (nuclei, blue) after the treatment of culture medium, 1p, 2p, 3p, 4p, 5p, and 6p for 12 h. Scale bars = 20 μm.

The addition of TNAP inhibitor (DQB) reduces the formation of nanofibers of 1 on Saos-2 cells (Figure S31, SI) and lessens the disruption of cytoskeletons (Figures S33 and 34, SI), further supporting the critical role of TNAP for EISA in inducing Saos-2 cell death. In addition, increasing the concentration of DQB reduces nanofiber formation (Figure S31, SI), suggesting that the inhibition of dephosphorylation depends on the concentration of DQB, which is supported by the higher concentrations of DQB rescuing more Saos-2 cells coincubated with 1p (Figure S32, SI). Incubating HS-5 cells with the self-assembling molecule 1, we examine the effect of hydrogelator on the normal cells. The Congo red staining results reveal that the assemblies of 1 hardly accumulate on HS-5 cells, even at the concentration of 100 μM. (Figure S35, SI) Moreover, 1 exhibits little effect on the cytoskeletons of HS-5 cells, at concentrations both below (i.e., 2 μM) and above (i.e., 10 μM) its cmc (2.7 μM) (Figures S36 and 37, SI). Only when the concentration is high enough (i.e., 100 μM), 1 starts to show slight disruptions on the cytoskeletons of HS-5 cells (Figures S36 and 37, SI).

Conclusion

In conclusion, this study establishes the self-assembling ability of EISA molecules as a key thermodynamic parameter for determining the efficacy of EISA against cancer cells that overexpresses certain enzymes. As revealed by the time-dependent dephosphorylation experiment (Figure S38, SI), the dephosphorylation rate of precursor 1p largely depends on the concentration of enzymes. This result indicates that, while the self-assembling ability determines the thermodynamic properties of EISA, the enzyme expression (and genetic information[50]) of cancer cells kinetically controls the EISA process. Taken together, these two parameters suggest that it is feasible to obtain the thermodynamic and kinetic properties of the EISA molecules in cell-free assays to predict the efficacy of EISA. Moreover, the understanding of the molecular transformation of the precursors in cell milieu (e.g., Figure ) not only highlights the complexity of cells but also illustrates how multiple enzymes control the EISA process to enable precise regulation of the formation of the assemblies in different cellular environments. In fact, the rapid dephosphorylation results in quick building up of the assemblies, which likely hampers the hydrolysis of the diesters (Figure S39, SI). Such kinetic control may be particularly useful for selectively targeting the desired cancer cells,[51] thus greatly reducing the off-target effects of assemblies. Although in this work we demonstrate the use of EISA for inhibiting cancer cells, the insights obtained here should be applicable to other applications, like molecular imaging,[25,52] analyte detection,[53] and vaccine adjuvants,[6] as well as the understanding of cytotoxicity of pathogenic assemblies (e.g., β-amyloids).[54] Moreover, the recent advances in structural biology and cell biology have revealed that nature uses a dynamic continuum of protein assemblies to control cellular processes.[55] Thus, this work also underscores a thermodynamic aspect for developing a dynamic continuum of supramolecular nanostructures as a functional mimic of higher-order protein assemblies.