Phase separation of a heterochiral peptide-drug conjugate amplified by ionic interactions Authors: Yangkai Zhou, Jinmin Ye, Xingru Liu, Shijian Liang, Huipeng Ma, Xinghua Shi, Hui Wang, Jiayang Li, Chunying Chen Published: 2026 ABSTRACT Artificial membraneless organelles offer a platform for understanding biological compartmentalization and advancing biomedical applications, while designing them with precise chemical control remains challenging. Herein, inspired by natural heterochiral peptide systems, we have developed a chiral engineered tripeptide-drug conjugate with alternating D/L residues, which undergoes phase separation through a stereochemical-ionic interplay, enabling cation-tunable liquid–liquid phase separation (LLPS) under physiological conditions. Notably, metastable heterochiral condensates can be stabilized by Na⁺ or K⁺ via cation–tripeptide interactions. Specifically, heterochiral configuration possibly introduces a stereochemical effect that elevates the energy barrier for liquid-to-solid phase transition, redirecting assembly toward LLPS rather than fibrillization, whereas homochiral diastereomers preferentially form β-sheet-rich hydrogels. Further, chirality-controlled phase behavior results in altered drug properties, including cellular uptake and liver metabolism. By altering chirality of a single residue, we present a minimalist stereochemical strategy for synthetically controlling over the energy landscapes of peptide-based condensates, expanding functional versatility through rational design. Introduction Membraneless organelles (MLOs) are liquid-like functional condensates of biomolecules that feature a surface-tension-driven spherical shape and participation of non-covalent interactions, including hydrogen bonding; electrostatic, dipole-dipole, π-π, cation-π interactions, and hydrophobic effects1,2. MLOs participate in a broad range of biological events such as cell signaling3, biochemical catalysis4,5, and electrochemical equilibrium6. The ubiquity and utility of MLOs inspire researchers to construct artificial MLOs (AMLOs) as in vitro models to understand the molecular nature of MLOs and mimic the natural functions of MLOs. AMLOs have also been applied to biomedical fields, such as drug/biomacromolecule delivery7,8, hypoglycemic rescue9, and tumor therapies10. However, current approaches to produce AMLOs remain limited, including biotechnological approaches (e.g., protein engineering or protein-RNA interaction) and chemical approaches (e.g., peptide modification or charged polymers)4,10–12. For chemical approaches, additional modification of the molecular backbone or multiple components might be required to produce AMLOs. Understanding how subtle structural details, such as amino acid chirality and the resulting spatial configurations, influence the assembly process and condensate properties is crucial for deciphering their complex biological roles and dysfunctions. These underscore the urgency to expand molecular engineering strategies for AMLO fabrication. Chiral molecules exist widely in the biological system (e.g., amino acids, sugars, nucleotides). Inspired by natural chirality, chiral materials have been developed and applied to biomedical fields13–15. The intrinsic chirality of the materials modulates their biological effects, including ligand-receptor recognition16, DNA binding17, and enzymic degradation18. Peptide-based materials, such as peptide-drug conjugates (PDCs), provide a chiral amino acid backbone that could be easily engineered to produce various chiral materials. In mammals, peptides and proteins exhibit virtually complete homochirality of L-amino acids. However, D-amino acid-containing heterochiral peptides have also been detected, such as dermorphin (opioid receptor agonist), achatina (neuropeptide), and bacitracin (antibiotic)19. Related studies have indicated the essential role of doped D-amino acids in regulating biological functions of these heterochiral peptides. For example, the replacement of D-Ala with L-Ala in dermorphin results in a significant reduction in opioid activity20. Therefore, tuning the chirality of peptides could be a promising strategy to regulate the properties of PDCs. Whereas most strategies tune the chirality of peptides from a homochiral L-configuration to a homochiral D-configuration, the introduction of heterochirality to peptides is less explored. Ionic interactions play pivotal roles in maintaining protein structure and biological activity. The importance of ions in biological condensates is increasingly recognized; however, the molecular insights into how ions influence condensate formation through modulating chemical spatial configurations, particularly in conjunction with chirality, are largely unknown. Studies reveal that specific tripeptide or tetrapeptide motifs, such as the “niche” structure, adopt conformations stabilized by cations (e.g., Na⁺, K⁺, Ca²⁺) via electrostatic interactions with mainchain carbonyl groups, enabling critical functions like ion transport and enzyme activation21. These motifs, often located at α-helix termini, demonstrate how peptide backbone geometry—governed by φ/ψ angles—facilitates ion coordination. Introducing heterochirality (D/L-amino acids) into short peptides enhances structural versatility22, where the chirality of amino acids may influence ion-binding capacity. D-amino acids disrupt canonical secondary structures (e.g., α-helices) but stabilize alternative conformations, such as β-turns, altering ion accessibility23,24. For instance, heterochiral peptides incorporating D-residues exhibit enhanced proteolytic resistance and modified ion-binding properties, enabling applications in antimicrobial peptide design25,26 and metal-ion chelation27. Recent work highlights that sodium ions, in particular, promote structural reorganization in heterochiral peptides, enhancing their stability and functional versatility28,29. These insights underscore the interplay between peptide chirality, ionic interactions, and functional nanostructures in biological systems. So far, chemical parameters such as pH, polarity, and ionic strength have dominated LLPS research, but chirality, a defining feature of biomolecules, has been largely overlooked. The role of chirality as a precise chemical handle for designing and controlling LLPS remains largely underexplored. In this work, inspired by the stereochemical adaptability of natural heterochiral peptides, we have employed a chiral engineering strategy to regulate the drug behavior of tripeptide-drug conjugates, leveraging the advantages of tripeptide design to mimic protein-like niche architectures. Herein, we have implemented a chiral programming on [2-naphthaleneacetic acid-Phe-Phe-Lys(benzenesulfonamide)-OH] (N-pepABS), a hypoxia-targeting self-assembled PDC that we previously reported30, to produce four pairs of PDC enantiomers, i.e., DDD/LLL, LDD/DLL, DLD/LDL, DDL/LLD, representing the chirality sequence of Phe-Phe-Lys tripeptide backbone. Whereas most chiral PDCs form gels under physiological conditions, a certain pair of heterochiral PDCs with alternating chirality (DLD/LDL) is observed to form liquid–liquid phase separation (LLPS). The presence of ionic species, including Na+ ions and negatively charged DLD species, is revealed in the coacervate phase. In addition, the interaction between Na+ ions or K+ (not Li+, Rb+, or Cs+) and carbonyl & carboxyl groups can produce a “niche-like” structure, and thus inhibits the formation of amyloid-like nanofibers. Investigation on the energy landscape of all isomers of PDC further reveals that chiral engineering produced stereochemical effects both on the energy barrier and thermostability of the liquid-to-gel transition. Of note, heterochirality of drug-conjugated Lys and adjacent Phe can increase the energy barrier for liquid-to-gel transition, probably owing to drug conjugation-produced steric hindrance, while homochirality of two adjacent Phe corresponds with better thermostability. Finally, this chirality-induced LLPS confers significant advantages for drug delivery with enhanced cell internalization, endowing inhibition of tumor proliferation in vivo. In summary, our results have proven chiral engineering as a promising strategy for kinetic and thermodynamic control over peptide-based condensates and regulate drug behavior of PDCs, expanding future drug design and discovery into heterochiral mirror space. (Fig. 1).Fig. 1Phase separation of a heterochiral peptide-drug conjugate amplified by ionic interactions.Chiral engineering of a tripeptide-drug conjugate, with alternating D/L residues (DLD), altered the energy barriers for liquid-to-gel transition and resulted in phase separation through stereochemical-ionic interplay. Under physiological conditions, where Na⁺ or K⁺ stabilize metastable condensates, certain heterochirality may introduce a stereochemical lock that expedites assembly toward LLPS rather than fibrillization, while other diastereomers prefer forming β-sheet-rich hydrogels. This has established a chiral engineering strategy to modulate drug properties. DLD exhibits faster cell uptake efficiency than DDD, while DDD is metabolized faster in the liver. Image(s) are provided by Servier Medical Art (https://smart.servier.com), licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Phase separation of a heterochiral peptide-drug conjugate amplified by ionic interactions. Chiral engineering of a tripeptide-drug conjugate, with alternating D/L residues (DLD), altered the energy barriers for liquid-to-gel transition and resulted in phase separation through stereochemical-ionic interplay. Under physiological conditions, where Na⁺ or K⁺ stabilize metastable condensates, certain heterochirality may introduce a stereochemical lock that expedites assembly toward LLPS rather than fibrillization, while other diastereomers prefer forming β-sheet-rich hydrogels. This has established a chiral engineering strategy to modulate drug properties. DLD exhibits faster cell uptake efficiency than DDD, while DDD is metabolized faster in the liver. Image(s) are provided by Servier Medical Art (https://smart.servier.com), licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Results Chiral engineering produces liquid–liquid phase separation To investigate chiral effects on self-assembling short peptides, we performed chiral engineering on N-pepABS, a hypoxia-targeting PDC that could self-assemble into amyloid-like nanofibers on the cell surface, as we previously reported. Stereochemical reprogramming of three α-carbon chiral centers in the Phe-Phe-Lys backbone generates four PDC enantiomer pairs, i.e., DDD/LLL, LDD/DLL, DLD/LDL, DDL/LLD, representing the chirality sequence of the Phe-Phe-Lys tripeptide backbone (Fig. 2a). Enantiomeric pairs of N-pepABS exhibited similar nuclear magnetic resonance (NMR) spectrums and symmetric circular dichroism (CD) pattern (Supplementary Figs. 1–5). The phase behavior of these peptides (1 mM) was first investigated under physiological conditions (pH 7.5 & 150 mM NaCl). It has been observed that homochiral pairs DDD/LLL and two heterochiral pairs (LDD/DLL and DDL/LLD) may form β-sheet-rich hydrogels through amyloid-like nanofibers entanglement (Fig. 2b). However, heterochiral DLD/LDL with alternating chirality produced a turbid solution, which is distinct from the other three groups (Fig. 2b, c). Microscopic imaging revealed the presence of micro-sized droplets, indicating the heterochiral DLD/LDL pair resisted fibrillization to undergo LLPS-driven condensate formation (Fig. 2b). Doping of 4-nitrobenzo-2-oxa-1,3-diazole labeled DLD (NBD-DLD) proved the distribution of the heterochiral PDC within the droplets (Supplementary Fig. 6). A recent study developed a label-free detection method for condensates with the second harmonic generation (SHG) technique, in which two photons (with frequency ω) interact with a non-centrosymmetric medium to produce one photon with a frequency of 2ω31. A SHG signal at around 518 nm was detected when DLD droplets were excited with a femtosecond laser (1037 nm), which could derive from the non-centrosymmetric liquid–liquid interface (Supplementary Fig. 7).Fig. 2Chiral engineering induces liquid–liquid phase separation.a A schematic illustration of the varied phase behavior of chiral-engineered N-pepABS. b Representative microscopic images and gelation properties (insert) of chiral-engineered DLD at pH 7.5 and 150 mM NaCl. Scale bar, 50 μm. Three independent experiments were performed. c Turbidity of chiral molecules at pH 7.5 and 150 mM NaCl measured by OD600. Data represent mean ± SD for n = 3 measurements. d Time-lapse microscopy of DLD at pH 7.5 and 150 mM NaCl. Scale bar, 20 μm. e Measurement of fluorescence recovery after photobleaching via confocal microscopy (up) and the corresponding fluorescence intensity (down). Scale bar, 5 μm. Three independent experiments were performed. Representative fluorescent images and fluorescent recovery curve (presented by mean ± SD for n = 3 measurements) were given. Part of image(s) are provided by Servier Medical Art (https://smart.servier.com), licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/) (a). Source Data are provided with this paper. Chiral engineering induces liquid–liquid phase separation. a A schematic illustration of the varied phase behavior of chiral-engineered N-pepABS. b Representative microscopic images and gelation properties (insert) of chiral-engineered DLD at pH 7.5 and 150 mM NaCl. Scale bar, 50 μm. Three independent experiments were performed. c Turbidity of chiral molecules at pH 7.5 and 150 mM NaCl measured by OD600. Data represent mean ± SD for n = 3 measurements. d Time-lapse microscopy of DLD at pH 7.5 and 150 mM NaCl. Scale bar, 20 μm. e Measurement of fluorescence recovery after photobleaching via confocal microscopy (up) and the corresponding fluorescence intensity (down). Scale bar, 5 μm. Three independent experiments were performed. Representative fluorescent images and fluorescent recovery curve (presented by mean ± SD for n = 3 measurements) were given. Part of image(s) are provided by Servier Medical Art (https://smart.servier.com), licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/) (a). Source Data are provided with this paper. The intrinsic fluidity of phase-separated droplets, characterized by their dynamic molecular exchange and interfacial adaptability, serves as a foundational feature for microreactor functionality, enabling continuous influx/efflux of substrates and products. The DLD droplets we obtained do exhibit liquid-like fluidity as illustrated by the fluorescence recovery after photobleaching (FRAP) analysis and droplet fusion (Fig. 2d, e). Similarly, fluidity was also observed for LDL enantiomer (Supplementary Fig. 8). Maintaining the fluidity of phase-separated droplets can be very important to function as microreactors by increasing the partition of hydrophobic enzyme or catalyst inside the droplets32. Herein, we validated the potential of such heterochiral droplets to become functional AMLOs with considerable fluidity. β-Gal was selected as a model enzyme to test its enzymic activity through X-Gal assay, where the substrate 5-Bromo-4-chloro-3-indolyl β-D-galactoside (X-Gal) was digested by β-Gal to produce blue products, which could be monitored through UV-Vis spectrum (Supplementary Fig. 9a). The characteristic peak at 620 nm increased more rapidly in the presence of LLPS (Supplementary Fig. 9b–d). No significant difference between DLD and LDL was observed (Supplementary Fig. 10). The underlying mechanism is associated with the properties of liquid-like condensates, which provide a distinct microenvironment capable of sequestering and concentrating a variety of biological molecules, while allowing the exchange of substrates and products33. Fluorescent imaging confirmed the co-location of heterochiral DLD and β-Gal inside the droplets, indicating the LLPS could efficiently recruit β-Gal to increase the local enzymic concentration (Supplementary Fig. 9e, f). Together, these results indicated the formation of LLPS under physiological conditions for DLD/LDL but not for other diastereoisomers. Asymmetric ion distribution revealed ionic species in heterochiral coacervate The dramatic effect of heterochiral engineering encouraged us to further investigate the mechanism of chiral-induced phase separation. Ionic interaction has been widely reported to be a key issue in regulating the peptide folding34,35, which also plays an essential role in LLPS36. Interestingly, it seems that the chirality of the peptide mainchain can afford more distinctive conformation under the influence of ionic interactions29. To better understand the phase behavior of chiral-induced phase separation, here we have constructed a phase diagram to study the ionic interactions within heterochiral DLD. As shown in Fig. 3b, it has been observed that the increased amount of NaCl can extend the concentration and pH range for LLPS of DLD. However, although ionic interactions can expand the operational window for LLPS by enhancing phase stability under diverse conditions, they remain incapable of overriding the pH 6.5 threshold, which strictly defines the forbidden regime for LLPS initiation. For example, droplets can be observed at pH 6.5 and 7.5, while gel-like precipitate would appear under the condition of pH 5.5 (Supplementary Fig. 11).Fig. 3Asymmetric ion distribution revealed ionic species in heterochiral coacervate.a A schematic illustration of the composition of DLD coacervate. b Phase diagram of DLD at two different NaCl concentrations. c Fraction of free monomer compared with the theoretical fraction of ionized species (obtained from pKa value). Data represent mean ± SD for n = 3 measurements. X-ray photoelectron spectrum of DLD coacervates for S 2p (d), Na 1 s (e), and Cl 2p (f). g Density functional theory (DFT) simulation of possible interactions between ionized DLD− and Na+. The geometry and binding free energy at B3LYP + D3(BJ)/def2TZVP are given. Note that the given free energy includes both ion binding and peptide conformation change. Part of image(s) are provided by Servier Medical Art (https://smart.servier.com), licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/) (a). Source Data are provided with this paper. Asymmetric ion distribution revealed ionic species in heterochiral coacervate. a A schematic illustration of the composition of DLD coacervate. b Phase diagram of DLD at two different NaCl concentrations. c Fraction of free monomer compared with the theoretical fraction of ionized species (obtained from pKa value). Data represent mean ± SD for n = 3 measurements. X-ray photoelectron spectrum of DLD coacervates for S 2p (d), Na 1 s (e), and Cl 2p (f). g Density functional theory (DFT) simulation of possible interactions between ionized DLD− and Na+. The geometry and binding free energy at B3LYP + D3(BJ)/def2TZVP are given. Note that the given free energy includes both ion binding and peptide conformation change. Part of image(s) are provided by Servier Medical Art (https://smart.servier.com), licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/) (a). Source Data are provided with this paper. It was then hypothesized that negatively charged DLD species dominated the phase behavior through Na⁺-mediated interaction to carboxylate/sulfonamide motifs from its heterochiral backbone. The titration of DLD has revealed its weak acid behavior with pKa ≈6.5 (carboxyl) and pKa2 ≈ 10.2 (sulfonamide), confirming predominant ionization of DLD at pH 7.5 (Supplementary Fig. 12). The carboxyl pKa has shown a significant upshift compared to free lysine (2.16), which could be a result of hydrophobicity-induced pKa shift37. Therefore, the majority of DLD was considered ionized at pH 7.5. The measurement of free monomer fraction under various ionic strength conditions can help us to obtain more details of the species distribution in DLD. As shown in Fig. 3c, under a NaCl concentration of 50 mM, where no LLPS occurs, the fraction of free monomer at the corresponding pH is close to the theoretical curve of the ionized species. However, increased ion concentration results in considerable deviation from the curve, indicating that more DLD has participated in the phase transition. Such a phenomenon is ascribed to either the presence of ionized species inside the coacervate phase or an upshifted pKa at higher ionic strength. The latter possibility is excluded when DLD was titrated under various ionic strengths (Supplementary Fig. 12). Accordingly, we also monitored the protonation state of COOH by FTIR. In the solution phase, the band at 1721 cm−1 (pH 2) is ascribed to the stretching mode of C=O for protonated COOH, while the band at 1590 cm−1 (pH 9) is ascribed to the asymmetric stretching mode of COO−. At pH 9, NaCl was further added to induce LLPS of DLD. As the majority of DLD was distributed in dense phase (Fig. 3c), we suggest that FTIR signals were derived from phase-separated DLD in dense phase instead of free monomers. It was observed that the FTIR spectrum of DLD in dense phase was similar to that of free DLD at pH 9, which indicates the presence of ionized species in DLD condensates (Supplementary Fig. 13). Collectively, these results suggested that NaCl could recruit ionized DLD species in the coacervate. To gain direct insight into the composition of the heterochiral coacervate, we have then performed elementary analysis with X-ray photoelectron spectroscopy (XPS). The coacervate phase of DLD at pH 7.5 was collected by centrifugation, washed gently with distilled water, and lyophilized for XPS analysis. The XPS spectrum exhibits obvious signals for S and Na, but weak signals for Cl (Fig. 3d–f). The relative element ratio of S, Na, and Cl is calculated to be 75.08%, 24.57%, and 1.35%, respectively. The molar ratio of DLD and Na, determined from the relative abundance of S and Na, is about 1.5:1, indicating the presence of 65.5% of ionized species in the coacervate of DLD. The above results revealed the presence of ionized species in the DLD dense phase. Therefore, we next explored the potential interaction between Na+ and ionized DLD with Density Functional Theory (DFT). Among various interaction modes, interaction between Na+ and carbonyl & carboxyl groups exhibits the lowest energy (Fig. 3g), indicating Na+ induced conformation change for DLD. To validate our simulation, FTIR spectra of DLD at various NaCl concentrations were recorded. In accordance with our simulation, increased NaCl concentration produces a slight red shift of C=O stretching vibration from 1647 to 1637 cm−1 (Supplementary Fig. 14). Further, we conducted NMR titration of DLD in aqueous solution and observed shifts in aromatic region in a salt-dependent-manner, suggesting a salt-dependent conformation change (Supplementary Fig. 15). Collectively, these findings establish that heterochiral coacervates are structurally governed by Na⁺-mediated ionic interaction between ionized DLD motifs (Fig. 3a). Specifically, pH-triggered protonation of carboxyl groups within the coacervate phase (at pH 6.5) induces liquid-to-gel transitions through charge shielding (Supplementary Fig. 16). DFT simulation further revealed Na+ induced conformation change, which could possibly be vital for the formation of DLD condensates. Ion promotes phase separation via inhibition of fibrilization We further investigated the driving force for LLPS of DLD. Of note, heating of the sample disrupts phase separation, which can be reproduced during cooling, featuring an upper critical solution temperature (UCST), which is similar to phase separation of resilin-like polypeptide (ELPs) that is primarily governed by hydrophobic interactions (Supplementary Fig. 17). To test this, DLD droplets were treated with 1,6-hexanediol (1,6-HD), a hydrophobic interaction disruptor1. 1% of 1,6-HD significantly disrupts the LLPS formed by DLD, as described by decreased turbidity (Supplementary Fig. 16a). Therefore, hydrophobic interaction, which is possibly derived from unionized species of DLD, could be the driving force of LLPS. Meanwhile, the lack of unionized species could explain why LLPS was not observed at higher pH values for DLD (Fig. 3b and Supplementary Fig. 11). Therefore, LLPS of DLD is driven by hydrophobic interactions. This corresponds with the salting-out effect, reducing their solubility and promoting peptide aggregation, thereby facilitating LLPS. However, the simple “salting-out effect” could not explain the pH dependence of DLD LLPS and the asymmetric distribution of Na+ and Cl+ across the liquid–liquid interface. Hence, the role of cations in DLD LLPS should be further investigated. Therefore, we investigated the influence of NaCl concentration on the morphology of DLD. At pH 7.5, different concentrations of NaCl were introduced to 1 mM of DLD and left for four hours. Transmission electron microscope (TEM) reveals that at a low ionic strength (13 mM), DLD forms amyloid-like nanofibers, indicating formation of a gel just as the other chiral peptides (Figs. 2b and 4c). As the ionic strength increased, the fraction of nanofibers decreased while a higher abundance of nanospheres was observed. At 100 mM NaCl, uniformly distributed nanospheres were observed, exhibiting a narrow size distribution with a low polymer dispersity index (PDI) measured by dynamic light scattering (DLS). Upon reaching 150 mM NaCl, macroscopic LLPS emerges, concomitant with the coalescence of nanoscale spheres, which corresponded with DLS results (Fig. 3b, c). Thioflavin T (ThT), a β-sheet sensitive fluorescent probe, has also been applied to investigate the formation of amyloid-like nanofibers38. In accordance with these results, the ThT kinetic assay at 13 mM NaCl exhibits a typical sigmoidal curve, while increased NaCl concentration significantly suppresses the growth of ThT fluorescence (Supplementary Fig. 19). Energy dispersive X-ray spectroscopy (EDX) mapping was applied to visualize the elemental distribution of different nanostructures. Correspondingly, the element mapping of a DLD nanospheres reveals the presence of Na element, indicating the presence of Na+ ions and ionized species in the nanospheres (Fig. 4e). In contrast, for nanofibers, the signal from Na element is close to the background level (Fig. 4d). These findings demonstrate that NaCl orchestrates the supramolecular assembly of DLD at pH 7.5, controlling the formation of β-sheet-dominated fibrils and LLPS condensates via ionic interaction, which can be mechanistically validated by DFT simulations (Fig. 4f). Of note, Na+ interacts with carbonyl & carboxyl group, forming a “niche-like” structure, in which mainchain carbonyl groups from three to four residues might be bridged by cations or cationic groups21. Formation of such a niche-like structure twists the peptide backbone and produces a disordered peptide bond arrangement, which might be the reason for β-sheet inhibition that requires an ordered arrangement of peptide bonds for hydrogen bond formation. Together, these results suggested NaCl could inhibit the formation of β-sheet-rich nanofibers and promote LLPS of DLD, probably through interaction between Na+ and charged peptide backbone, which induced conformation change that is not favorable for β-sheet formation (Fig. 4a).Fig. 4Ion promotes phase separation via inhibition of fibrilization.a A schematic illustration of NaCl-dependent phase separation of DLD. b Size distribution of assembled DLD at various NaCl concentrations measured by DLS. c Representative transmission electron microscope (TEM) images of DLD dispersed at various NaCl concentrations (pH 7.5). Samples were prepared 4 h after dispersion. Scale bar, 200 nm. Three independent experiments were performed. High-angle annular dark-field (HAADF) scanning TEM images and elemental mappings of nanofiber (d) and nanosphere (e) formed by DLD at various NaCl concentrations. Scale bar, 20 nm. Three independent experiments were performed. f Conformations of DLD with or without Na+ binding calculated by density functional theory (DFT), producing a niche-like structure. Part of image(s) are provided by Servier Medical Art (https://smart.servier.com), licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/) (a). Source Data are provided with this paper. Ion promotes phase separation via inhibition of fibrilization. a A schematic illustration of NaCl-dependent phase separation of DLD. b Size distribution of assembled DLD at various NaCl concentrations measured by DLS. c Representative transmission electron microscope (TEM) images of DLD dispersed at various NaCl concentrations (pH 7.5). Samples were prepared 4 h after dispersion. Scale bar, 200 nm. Three independent experiments were performed. High-angle annular dark-field (HAADF) scanning TEM images and elemental mappings of nanofiber (d) and nanosphere (e) formed by DLD at various NaCl concentrations. Scale bar, 20 nm. Three independent experiments were performed. f Conformations of DLD with or without Na+ binding calculated by density functional theory (DFT), producing a niche-like structure. Part of image(s) are provided by Servier Medical Art (https://smart.servier.com), licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/) (a). Source Data are provided with this paper. Monovalent alkali ions gate phase separation via radius control So far, our current data has demonstrated that niche-like motifs formed by the interplay between chiral peptide mainchains and sodium ions are essential for maintaining LLPS. To further probe it was a cation, but not an anion-mediated conformation change that altered supramolecular assembly, we systematically investigated alternative alkali metal ions (Li⁺, K⁺, Rb⁺, Cs⁺) with varying ionic radii. In the above experiments, the DLD stock solution was first prepared in NaOH before LLPS induction. To exclude the interference of Na+, the stock solution was prepared in DMSO instead, which was further dispersed into different salt buffer solutions (pH 7.5). Dispersion of the stock solution into all buffers results in enhanced turbidity, indicating that LLPS was rapidly driven upon mixing. However, centrifugation of the dispersions in LiCl, RbCl or CsCl produced white precipitates, while the dispersions in NaCl or KCl always kept their semi-transparent condensate phases (Supplementary Fig. 20a). Heating of all the five dispersions resulted in a homogenous transparent solution, while cooling of the solutions containing NaCl or KCl, but not LiCl, RbCl or CsCl, produced phase-separated droplet, accompanied with reproduced turbidity (Fig. 5b, c and Supplementary Fig. 21a). These results suggest that DLD could probably undergo LLPS in buffers of all 5 salts, driven by hydrophobic interactions, while cations might influence the stability of condensates. Different cations could exhibit varied capacity for maintaining LLPS.Fig. 5Monovalent alkali ions gate phase separation via radius control.a A schematic illustration of the effect of monovalent alkaline metal Ions on phase separation. b Representative microscopic images of DLD dispersed in different salt buffers (salt concentration: 150 mM, pH 7.5). Scale bar, 100 μm. Three independent experiments were performed. c Turbidity of DLD dispersed in different salt buffers measured by OD600. Data represent mean ± SD for n = 3 measurements. d Change of ThT fluorescence after DLD dispersed in different salt buffers for 12 h. Data represent mean ± SD for n = 3 measurements. DFT simulations of interaction between DLD and different alkaline metal ions (f) at B3LYP + D3(BJ)/def2TZVP and corresponding cation-oxygen distance (e). Data represent mean ± SD for the distance between the cation and the four adjacent oxygen atoms. Part of image(s) are provided by Servier Medical Art (https://smart.servier.com), licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/) (a). Source Data are provided with this paper. Monovalent alkali ions gate phase separation via radius control. a A schematic illustration of the effect of monovalent alkaline metal Ions on phase separation. b Representative microscopic images of DLD dispersed in different salt buffers (salt concentration: 150 mM, pH 7.5). Scale bar, 100 μm. Three independent experiments were performed. c Turbidity of DLD dispersed in different salt buffers measured by OD600. Data represent mean ± SD for n = 3 measurements. d Change of ThT fluorescence after DLD dispersed in different salt buffers for 12 h. Data represent mean ± SD for n = 3 measurements. DFT simulations of interaction between DLD and different alkaline metal ions (f) at B3LYP + D3(BJ)/def2TZVP and corresponding cation-oxygen distance (e). Data represent mean ± SD for the distance between the cation and the four adjacent oxygen atoms. Part of image(s) are provided by Servier Medical Art (https://smart.servier.com), licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/) (a). Source Data are provided with this paper. To compare all five alkaline metal solutions, the dynamics of the dispersion were also investigated by monitoring the change of the turbidity (Supplementary Fig. 20b). For LiCl, RbCl, and CsCl, the turbidity rapidly decreased, reaching a slightly reduced OD value at equilibrium, which could be ascribed to the precipitation process. For NaCl and KCl, the turbidity attenuated gradually to a lower value, which was ascribed to the fusion of smaller droplets. The half-time for the attenuation followed the sequence Cs+ > Rb+ > Li+ > K+ > Na+. Correspondingly, an increased ThT fluorescence, indicating the formation of β-sheet, was observed for diluted samples prepared with Li+, Rb+, and Cs+, but not those prepared with Na+ and K+ (Fig. 5d and Supplementary Fig. 21b). To further qualitatively compare the interaction between ions and DLD, we monitored the temperature dependence of phase change in the presence of different ions. The buffers containing different salts were first heated to 60 °C prior to the dissolution of DLD to obtain a clear solution. The turbidity was monitored during the cooling of the samples. It was hypothesized that a stronger peptide-ion interaction could result in a higher critical temperature for phase transition. Accordingly, the order for critical temperature was determined as CsCl < LiCl < RbCl < KCl < NaCl, which corresponds with the charge densities of these ions in aqueous solution (Supplementary Fig. 22). This result indicates that interactions between ions and LLPS peptide were associated with charge densities of these ions. To further understand the interaction between alkaline metal ions and DLD, DFT simulations of the interaction between alkaline metal ions and ionized DLD (carbonyl & carboxyl groups) were conducted (Fig. 5f and Supplementary Fig. 21c). Of note, the average distance between cation and oxygen atoms followed the order of Li+ (1.98 Å) < Na+ (2.04 Å) < K+ (2.70 Å) < Rb+ (2.87 Å) < Cs+ (3.05 Å), corresponded with the order of atomic radius of alkaline metal ions (Fig. 5e). However, Li+ could not induce LLPS according to our experiment results. XPS analysis of the precipitates obtained from dispersion in LiCl exhibits a negligible Li signal (Supplementary Fig. 21d). Of note, the sequence of Cs+, Rb+, K+, and Na+ corresponds to an increasing charge density trend (Fig. 5a), while Li+ became the only exception, as it is known to be strongly hydrated in aqueous solutions39,40. The water molecules within the first hydration shell are tightly bound to Li+, which decreases its effective surface charge density to a similar level as Rb+. Collectively, these results suggest that a strong ionic interaction between cation and ionized DLD, modulated by alkali metal ionic radii through charge screening and stereochemical constraints, is vital for maintaining LLPS of DLD. Chirality-encoded energy landscapes in liquid-to-gel transition After revealing the mechanism of NaCl-induced LLPS for DLD, we next explore why other isomer pairs can not form LLPS under the same conditions. Considering the vital role of ionized species in the coacervate phase, the pKa value of all chiral PDCs was evaluated by the NMR method. The pKa values for the carboxy group of all isomer pairs were measured as 4.44 for DDD/LLL, 4.65 for LDD/DLL, 4.13 for DLD/LDL, and 3.95 for DDL /LLD, respectively (Supplementary Fig. 23). However, DDD could not form LLPS at higher pH or higher NaCl concentrations, indicating the dissociation properties did not account for chirality-induced LLPS (Supplementary Fig. 24). Of note, previous research suggests LLPS as an intermediate stage for the self-assembly of short peptides followed by β-sheet formation to produce nanofibers41. Similar liquid-to-gel transition is also observed for natural proteins, such as pathology-related tau protein in Alzheimer’s disease42, α-synuclein43, and intrinsic disordered proteins31. To prove LLPS as the intermediate stage, we have introduced tannic acid (TA), which can inhibit the formation of β-sheet structures probably through hindered aromatic interactions44. Nile red, a fluorescent indicator of structures formed through hydrophobic interactions, was introduced to label phase-separated droplets45. When TA was added, phase-separated droplets with red fluorescence were observed for all chiral PDCs, including DDD, LDD, and DDL (Fig. 6b), corresponding with increased turbidity and reduced ThT fluorescence (Fig. 6c, d). These results indicate that LLPS is the intermediate stage during the gelation process of N-pepABS, and chiral effects may occur during the liquid-to-gel transition. To compare properties of TA-induced chiral condensates, we utilized fluorescence lifetime imaging microscopy (FLIM) to visualize viscosity and polarity of chiral condensates doped with boron dipyrromethene fluorophore BODIPY (viscosity sensitive) or 7-sulfonamide benzoxadiazole fluorophore SBD (polarity sensitive)36. The condensates formed by TA with different chiralities all exhibited gel-like properties, with no significant differences in apparent viscosity or polarity (Supplementary Fig. 25), indicating chiral effects may occur during the liquid-to-gel transition but not the formation of LLPS.Fig. 6Chirality-encoded energy landscapes in liquid-to-gel transition.a Chiral effects on the energy landscape of N-pepABS gelation. b Representative fluorescent microscope images of chiral N-pepABS with or without tannic acid (pH 7.5 & 150 mM NaCl). Droplets were labeled with Nile Red. Scale bar, 10 μm. Three independent experiments were performed. c Turbidity of chiral N-pepABS with or without tannic acid measured by OD600. Data represent mean ± SD for n = 3 measurements. d Change of ThT fluorescence of chiral N-pepABS when tannic acid was added. Data represent mean ± SD for n = 3 measurements. e The ThT kinetic assay of chiral N-pepABS (pH 7.5 & 150 mM NaCl). Data represent mean ± SD for n = 3 measurements. f A DFT simulation revealing interference of one Na+ ion to two layers of β-sheet formed by DDD or DLD peptide backbone. The drug motif was not added to simplify the simulation. g Effect of Na+ and chirality on peptide β-sheet as revealed by density functional theory (DFT) simulation. h A chiral engineering stereochemically modulated tetrahedron cavity formed by four benzene rings. Part of image(s) are provided by Servier Medical Art (https://smart.servier.com), licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/) (a). Source Data are provided with this paper. Chirality-encoded energy landscapes in liquid-to-gel transition. a Chiral effects on the energy landscape of N-pepABS gelation. b Representative fluorescent microscope images of chiral N-pepABS with or without tannic acid (pH 7.5 & 150 mM NaCl). Droplets were labeled with Nile Red. Scale bar, 10 μm. Three independent experiments were performed. c Turbidity of chiral N-pepABS with or without tannic acid measured by OD600. Data represent mean ± SD for n = 3 measurements. d Change of ThT fluorescence of chiral N-pepABS when tannic acid was added. Data represent mean ± SD for n = 3 measurements. e The ThT kinetic assay of chiral N-pepABS (pH 7.5 & 150 mM NaCl). Data represent mean ± SD for n = 3 measurements. f A DFT simulation revealing interference of one Na+ ion to two layers of β-sheet formed by DDD or DLD peptide backbone. The drug motif was not added to simplify the simulation. g Effect of Na+ and chirality on peptide β-sheet as revealed by density functional theory (DFT) simulation. h A chiral engineering stereochemically modulated tetrahedron cavity formed by four benzene rings. Part of image(s) are provided by Servier Medical Art (https://smart.servier.com), licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/) (a). Source Data are provided with this paper. To explore the chiral effects during the liquid-to-gel transition, we first analyzed the energy barrier of the liquid-to-gel transition by kinetic ThT fluorescent assay at the same concentrations. Both DDD and LDD exhibited rapid growth of ThT fluorescence, followed by DDL and DLD (Fig. 6e). These results indicate that chiral PDCs exhibited different energy barriers for liquid-to-gel transition, in the order of DLD > DDL > LDD > DDD (Fig. 6a). This could be ascribed to the largest steric hindrance of Lys among the three amino acids owing to the conjugation of ABS, which required enhanced energy for conformation change to favor β-sheet formation. We next investigated the Gibbs free energy of monomer-gel transition, which can be obtained from free monomer concentration at equilibrium46:1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$c=\exp \left(-\frac{\triangle G}{{k}_{B}T}\right)$$\end{document}c=exp−△GkBTwhere \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$c$$\end{document}c is free monomer concentration obtained through NMR, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$T$$\end{document}T is the absolute temperature, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${k}_{B}$$\end{document}kB is the Boltzmann constant. The \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\triangle G$$\end{document}△G for DDD, LDD, DLD, and DDL is calculated to be −5.39 kcal/mol, −4.84 kcal/mol, −4.80 kcal/mol, and −4.99 kcal/mol, respectively (Fig. 6a and Supplementary Fig. 26). It is noted that molecules with a homochiral “Phe-Phe” sequence presented the largest \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\triangle {{{\rm{G}}}}$$\end{document}△G, indicating the homochirality of two adjacent Phe may contribute to the stability of amyloid-like nanofibers. To better understand hetreochirality produced stereochemical effect, we performed DFT simulation of a fiber-like structure formed by DDD and DLD. CD spectrum of DDD gel exhibits a negative peak at ~198 nm and a positive peak at ~223 nm, suggesting the presence of anti-parallel β-sheet structures (Supplementary Fig. 27). We further applied a Na+ ion between 2 layers of β-sheet formed by DDD or DLD peptide backbone (Fig. 6f). As indicated by changed orientation of C=O bonds, obvious torsion on peptide backbone conformation was observed when chiral editing was applied, suggesting disruption of hydrogen bonds (Fig. 6g). These results revealed that DDD β-sheet was more stable than that of DLD. To gain insights into the underlying chiral mechanism, we focus on the interaction between Na+ and the PDC β-sheet. Of note, Na+ was surrounded by four benzene rings from the side chain of Phe, which formed a tetrahedron cavity, indicating potential cation-π interactions. For DDD, the distance between Na+ and benzene rings ranged from 2.868 to 3.529 Å. For DLD, the location of one benzene ring was altered owing to heterochirality, which resulted in a changed cation-π interaction. It was observed that Na+ was closer to two of the benzene rings (2.763 Å and 2.819 Å), which led to deformation of the tetrahedron cavity formed by 4 benzene rings (Fig. 6h). As a result, the β-sheet of DLD was tortured owing to altered chirality and cation-π interaction. Collectively, our results have revealed that N-pepABS forms a gel through an intermediate stage of LLPS, while chiral engineering produced chiral effects on the energy landscape of liquid-to-gel transition (Fig. 6a). Especially, heterochiral DLD/LDL with alternating chirality produced the largest energy barrier for liquid-to-gel transition, resulting in a different phase behavior under physiological conditions compared to other isomer pairs. Of note, we found that the DLD peptide without drug conjugation could not form LLPS droplets, probably owing to decreased hydrophobicity and side-chain steric hindrance (Supplementary Fig. 28). Chiral engineering altered drug properties Our results have demonstrated the possibility of regulating PDC properties in physiological conditions by adjusting the chirality of the peptide backbone. To explore the chiral effect at the biological interface, we first compared the enzyme stability of all 8 chiral isomers using the protease K assay. While most isomers exhibited resistance to protease K, complete digestion of LLL and partial digestion of LLD was observed. These results indicate that incorporation of one D-amino acid could increase enzyme resistance of PDCs (Supplementary Fig. 29). We next investigated states of DLD and DDD in the presence of fetal bovine serum (FBS). It was reported that condensates could be destroyed by serum proteins at high concentration47. Similarly, it was observed that the addition of fetal bovine serum (FBS) significantly reduced the turbidity of the sample, indicating rupture of DLD condensates. We further monitor such a process with fluorescence correlation spectroscopy (FCS), where FITC-labeled DLD was doped to characterize the diffusion behavior of DLD in the presence of FBS. It was found that DLD condensates were transformed into nanoaggregates post-FBS dilution with an average size of ~4 nm (Supplementary Fig. 30a–c). Moreover, we also investigated the effect of serum on DDD. In the presence of serum, a gel-like precipitate could be observed after centrifugation, indicating DDD nanofibers were stable enough to tolerate serum (Supplementary Fig. 30d, e). These results also indicate that DLD condensates were less stable than DDD fibers, as phase separation of DLD was considered as a metastable intermediate stage during liquid-to-gel transition (Fig. 7a).Fig. 7Chiral engineering altered drug properties.a A state of DLD and DDD in the presence of serum. Cell viability of MDA-MB-231 cells incubated with different concentrations of DLD (b) or DDD (c) for 24 h under hypoxic conditions. Data represent mean ± SD for n = 6 biological replicates. d Fluorescent images of MDA-MB-231 cells incubated with DLD (labeled with NBD-DLD) or DDD (labeled with NBD-DDD) at different time points. Scale bar, 50 μm. Data represent mean ± SEM for n = 20 cells at different views. e Comparison of anti-tumoral effect between DDD and DLD. f Tumor growth curves of 4T1-bearing mice with various treatments. Data represent mean ± SD for n = 5 biological replicates. Statistical significance was calculated using a two-tailed Student’s t-test. Biodistribution of DLD (g) and DDD (h) at various time points post i.v. injection. Data are presented as mean values ± SD for n = 3 biological replicates. Part of image(s) are provided by Servier Medical Art (https://smart.servier.com), licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/) (a, e). Source Data are provided with this paper. Chiral engineering altered drug properties. a A state of DLD and DDD in the presence of serum. Cell viability of MDA-MB-231 cells incubated with different concentrations of DLD (b) or DDD (c) for 24 h under hypoxic conditions. Data represent mean ± SD for n = 6 biological replicates. d Fluorescent images of MDA-MB-231 cells incubated with DLD (labeled with NBD-DLD) or DDD (labeled with NBD-DDD) at different time points. Scale bar, 50 μm. Data represent mean ± SEM for n = 20 cells at different views. e Comparison of anti-tumoral effect between DDD and DLD. f Tumor growth curves of 4T1-bearing mice with various treatments. Data represent mean ± SD for n = 5 biological replicates. Statistical significance was calculated using a two-tailed Student’s t-test. Biodistribution of DLD (g) and DDD (h) at various time points post i.v. injection. Data are presented as mean values ± SD for n = 3 biological replicates. Part of image(s) are provided by Servier Medical Art (https://smart.servier.com), licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/) (a, e). Source Data are provided with this paper. Such a difference could possibly result in a distinct biological effect. As a peptide-drug conjugate, the original design of homochiral DDD targeted carbonic anhydrase IX (CAIX), which is highly expressed in hypoxic tumors. We next evaluated the cytotoxicity of homochiral DDD and heterochiral DLD on MDA-MB-231 tumor cells that highly express CAIX under hypoxic conditions. Our previous study on DDD showed that cytotoxicity was observed after incubation for 48 h. However, at 24 h, cell viability was not affected for DDD, while cytotoxicity was observed for DLD (Fig. 7b). It was then hypothesized that chiral-engineered DLD favored more efficient cell internalization owing to altered phase behavior. To test this hypothesis, we next doped NBD-DDD or NBD-DLD for fluorescent labeling and monitored cellular fluorescence at various time points to track the interactions between chiral PDCs and cells. It was observed that DLD-treated tumor cells exhibited faster cellular fluorescent accumulation than DDD treatment, indicating more efficient internalization (Fig. 7d). We further investigated the cellular uptake mechanism of DLD by flow cytometry. Internalization of DLD was inhibited by low temperature and several endocytosis inhibitors, including Methyl-β-cyclodextrin (inhibitor for lipid rafts/cholesterol-enriched microdomains/caveolae) and Amiloride (inhibitor for macropinocytosis), indicating DLD nanoaggregate could be internalized through multiple pathways (Supplementary Fig. 31a). Considering DLD was conjugated with a CAIX-targeting drug, while CAIX, whose expression could be upregulated by hypoxia, is reported to facilitate caveolae-mediated endocytosis, we further investigated the effect of hypoxia on DLD uptake. It was further found that cellular uptake of DLD could be enhanced under hypoxia (Supplementary Fig. 31b). Therefore, cellular uptake of DLD nanoaggregates involved multiple endocytosis pathways, such as caveolae-mediated endocytosis and macropinocytosis, which could be further enhanced by hypoxia. Together, these results suggested chiral engineering could alter cell internalization of N-pepABS, and could be a potential strategy in enhancing the targeting capacity of PDCs. Encouraged by efficient in vitro CAIX inhibition to hypoxic tumor cells, we next investigate whether chiral-engineered DLD could suppress tumor growth in vivo through intravenous administration. Chiral-engineered DLD (with or without drug conjugation) was intravenously injected into MDA-MD-231 tumor-bearing mice, followed by tumor volume monitoring (Supplementary Fig. 32a). Compared to the control group and treatment of heterochiral peptide alone, efficient tumor suppression was observed post DLD injection (Supplementary Fig. 32b, c), which was ascribed to conjugation of hypoxia-targeting drug and successful intravenous delivery. Correspondingly, most severe histological damage was found in the tumor treated with DLD PDCs (Supplementary Fig. 33a). Meanwhile, tumors treated with DLD PDCs also exhibited the lowest Ki67 level, suggesting suppressed tumor proliferation (Supplementary Fig. 33b). Correspondingly, TdT-mediated dUTP Nick-End Labeling (TUNEL) staining revealed obvious tumor apoptosis level post-treatment of DLD PDCs (Supplementary Fig. 33c). We further evaluated whether ABS-conjugated DLD could efficiently target and inhibit CAIX at tumor sites. Immunohistochemistry staining of tumor slices revealed significantly reduced CAIX level post-injections of DLD PDCs (Supplementary Fig. 33d), suggesting efficient CAIX inhibition. Of note, the body weight of DLD-treated mice remained steady without weight loss (Supplementary Fig. 32d), indicating good biocompatibility of chiral-engineered DLD. We further compared anti-tumoral efficiency in the 4T1 tumor model, where both DLD and DDD were intravenously injected (Fig. 7e). It was found that both DDD and DLD exhibited anti-tumoral effects during about 2-week treatment. Of note, DLD seemed to exhibit a better tumor inhibitory effect than DDD (Fig. 7f and Supplementary Fig. 34). The body weight of all mice remained steady during treatments (Supplementary Fig. 35). To understand such a difference, we compared the in vivo distribution of DLD and DDD. Liver accumulation of both PDCs was observed. Whereas DLD showed a maximum accumulation in the liver at about 2 h post-injection, DDD exhibited a rapid decrease from 0.5 to 2 h, indicating DDD was easily accumulated and metabolized in the liver (Fig. 7g, h). This could possibly be ascribed to the size difference between DLD and DDD during circulation, where DLD possibly interacted with blood proteins to form ultrafine nanocomplexes that favor longer circulation, while DDD maintained larger nanofibers that were easily accumulated and metabolized in the liver. Similarly, rapid liver metabolism of D-peptide nanofibers was reported by several studies48. At tumor sites, we observed a maximum accumulation at around 4 h post-injection for DLD (Supplementary Fig. 36), while DDD exhibited a rapid decrease in fluorescent signal from 0.5 to 4 h post-injection (Fig. 7h), which possibly explained why DLD exhibited a better anti-tumoral effect. Collectively, simply altering one residue’s chirality in a tripeptide-drug conjugate (PDC) provides a suite of benefits for drug delivery, including effective targeting of hypoxic tumors, enhanced tumor cell internalization, and the potential for improved intravenous administration. Importantly, these results suggest that chiral engineering could be a promising and minimalistic design to modulate drug properties of PDCs, probably associated with altered phase behaviors. Such a strategy challenges the notion that complex modifications are required for functional PDCs, offering a blueprint for repurposing existing peptide therapeutics through stereochemical editing. Discussion In summary, we have performed chiral engineering on a model tripeptide PDC (N-pepABS) to produce four pairs of chiral molecules. Formation of LLPS has been observed for a pair of heterochiral PDCs (DLD/LDL) with alternating chirality, while other chiral PDCs formed a gel. Investigation on the energy landscape of chiral PDCs further revealed that all chiral N-pepABS can form a hydrogel through an intermediate stage of LLPS, while chiral engineering produced chiral effects on the energy landscape of the liquid-to-gel transition. Especially, chiral engineering has produced a stereochemical effect on side chains that hinders β-sheet formation. Mimicking natural ion-binding proteins through niche-like motifs, Na+ ions (or K+, but not Li+, Rb+, or Cs+) can interact with the carbonyl & carboxyl groups of ionized species to further inhibit β-sheet formation, thus amplify phase difference among chiral molecules at physiological conditions and expand functional versatility. Therefore, chiral engineering and ionic regulation could provide kinetic and thermodynamic control to peptide-based condensates. Finally, we demonstrated chiral engineering as a potential strategy to alter the properties of PDCs, as chiral-engineered DLD can induce more efficient cellular internalization than homochiral DDD in tumor cells. Our results have demonstrated that both hydrophobic interactions (which derive from unionized species) and interactions between cations and negatively charged species are necessary to induce LLPS of DLD. The presence of NaCl in the coacervate droplets has also been reported49. Our results have revealed the presence of Na+ but negligible Cl− in the coacervate phase of DLD. This can be a result of ion concentration gradients formed at the coacervate interphase31,50, which was washed during XPS sample preparation in our experiment. The effect of monovalent alkaline metal ions has also been evaluated. LLPS can only be induced by Na+ or K+, but not Li+, Rb+, or Cs+. Generally, the interaction between cation and DLD follows the order of atomic radius, except for Li+. A similar phenomenon has also been reported by several researches39, due to the tightly bound first hydration shell of Li+ that decreases its effective charge density in aqueous phase39,40. The hydration layer of ions should be considered in future coacervate simulations. Of note, recent studies suggest phase-separated coacervates could produce a surface potential from ionic gradients and regular cellular electrochemical equilibria6,50. Our results provide an ionic insight to help understand these functions of biomolecular condensates. The structural integrity and functional efficacy of proteins are fundamentally sustained through ionic coordination mechanisms. Nests or niches are recurring structural motifs critical for binding anions or cations, which are evolutionarily conserved and functionally vital. For example, nests create concavities where mainchain NH groups coordinate anionic atoms (e.g., phosphate, carboxylate), stabilize catalytic intermediates (e.g., oxyanion holes in proteases), are involved in P-loop activities (binding phosphate groups inside ATP or GTP-binding proteins), and bind Fe-S clusters (modulating the oxidation–reduction potential of the potentially catalytically active Fe atoms)51. On the other hand, cation-binding niches regulate ion channels and metal cofactors21. Research demonstrates that defined oligopeptide sequences, particularly tri- and tetrapeptides arranged in “Niche” configurations, achieve conformational stabilization through cation coordination (e.g., Na⁺, K⁺, Ca²⁺) mediated by electrostatic interactions with mainchain carbonyl groups. These cation-binding motifs, frequently observed at α-helical termini, exemplify how the precise spatial arrangement of peptide mainchains enables selective ion coordination to underlie essential biological processes, including ion transport and enzyme activation52. These structural motifs enable ligand stabilization, redox modulation, or phosphate-binding activation that offer novel strategies for synthetic biology. The tripeptide-based coacervates engineered to mimic protein-like niche/nest architectures might be precisely functionalized with catalytic or ion-channel functionalities, serving as a potent candidate of AMLOs. Importantly, strategic incorporation of heterochiral amino acid residues (combining L- and D-isomers) could expand the structural diversity of these motifs22. This stereochemical modification potentially modulates both ion-binding selectivity and the thermodynamic stability of formed complexes. The phenomenon highlights the intricate relationship between peptide backbone geometry and functional capacity in biological macromolecules. Therefore, our results have proven chiral engineering as a promising way to tune the phase behavior of PDCs and could be applicable to other types of PDCs. Certain heterochirality may introduce a stereochemical lock through cation–tripeptide coordination under physiological conditions, which elevates the energy barrier for β-sheet nucleation and induces assembly toward LLPS rather than fibrillization. Of note, DDD can rapidly form a hydrogel, that made it inappropriate for intravenous injection. Heterochiral DLD with alternating D/L chirality, in contrast, is possible for intravenous injection, which could expand its biomedical application. Delicate chiral engineering could expand future peptide screening into mirror space to obtain better specificity. As the smallest known niche-mimicking tripeptide, our system offers unparalleled synthetic accessibility. By preserving the drug’s structure while introducing heterochirality, we provide a blueprint for “chirality editing” of existing peptide therapeutics to enhance delivery. Our use of a simple tripeptide, representing a minimal “niche” unit, offers potential for rational drug design. By demonstrating that chirality can convert a known molecular scaffold into an LLPS-forming system, we propose a strategy to potentially enhance drug loading or transport capabilities of peptide-based carriers by transforming them into condensates via chiral engineering. However, PDCs with longer sequences may face difficulty in screening appropriate heterochiral sequences, as chiral combinations grow exponentially. Computational approaches or high-throughput screening techniques might be required to reduce the amount of experimental work in future studies. Methods Ethics statement This study complied with relevant ethical regulations and was approved by the ethics committee of the National Center for Nanoscience and Technology of China (approval numbers NCNST21-2208-0605 and NCNST21-202510-0125). Materials Fmoc-amino acids and resin were obtained from GL Biochem (Shanghai, China). Commercially available reagents were used without further purification. 4-(2-aminoethyl)-benzenesulfonamide, O-benzotriazole-N,N,N′,N′-tetramethyl-uronium-hexafluorophosphate (HBTU) and 1,1′-thiocarbonyldiimidazole (TCDI) were purchased from J&K. Organic Solvents including N,N′-dimethylformamide (DMF), N,N-Diisopropylethylamine (DIEA), methanol (MeOH), ethyl acetate, hexane, triethylamine, dichloromethane (DCM), ether and acetonitrile were purchased from Innochem. 4-chloro-7-nitro-2,1,3-benzoxadiazole (NBD-Cl) was purchased from Alfa Aesar. All chemicals were of reagent grade or better. The MDA-MB-231 cell line was purchased from the Procell. Dulbecco’s modified Eagle’s medium (DMEM), fetal bovine serum (FBS), and penicillin-streptomycin were purchased from Invitrogen Corporation. Cell Counting Kit-8 (CCK-8) was purchased from Beyotime Institute of Biotechnology (Shanghai, China). Characterization Nuclear magnetic resonance (NMR) spectra were recorded by AVANCE III HD 400. Mass spectrums (MS) were obtained by LCMS-8040 (Shimadzu) or Q-Exactive (Thermo Fisher). X-ray photoelectron spectra were recorded by ESCALAB 250Xi X. Transmission electron microscopy (TEM) imaging was carried out by Tecnai G2 20 S-TWIN. Elementary mapping was obtained by Tecnai G2 F20 U-TWIN. Microscopic images were captured with the Nikon Confocal Microscope System (AXR-NSPARC/Ji). Circular dichroism (CD) spectra were measured by JASCO J-1500. Synthesis of chiral short peptides (2-Naphthaleneacetic acid-Phe-Phe-Lys-OH, N-pep) All chiral short peptides were prepared via standard solid-phase peptide synthesis (SPPS) using 2-chlorotrityl chloride resin (0.8–1.3 mmol g−1) and the corresponding N-Fmoc protected amino acids. (Supplementary Fig. 37) In a typical synthesis, 1 g of 2-chlorotrityl chloride resin was fully swollen in anhydrous DCM for 15 min in a dry peptide synthesis tube. 2 mmol of first amino acid (Fmoc-L-Lys(Boc)-OH or Fmoc-D-Lys(Boc)-OH) was dissolved in 5 mL anhydrous DMF and 1 mL DIEA, which was further used for a 40 min coupling on resin. Then the resin was washed with DMF, followed by blocking of unreacted resin with a mixed solution (DCM:MeOH:DIEA = 17:2:1, v/v/v) for 30 min. After washing with DMF, the above resin was treated with piperidine (20% in DMF, v/v) for 30 min to remove Fmoc protection on amine groups. The resin was washed with DMF post-deprotection. 2 mmol of second amino acid (Fmoc-L-Phe-OH or Fmoc-D-Phe-OH) and 1.9 mmol of HBTU were mixed and dissolved in 5 mL anhydrous DMF and 1 mL DIEA, and added to deprotected resin. After coupling for 30 min, the resin was washed with DMF. Then, another two rounds of deprotection (in 20% piperidine) and coupling (Fmoc-L-Phe-OH or Fmoc-D-Phe-OH for the third coupling, and 2-naphthaleneacetic acid for the final coupling) were performed. After the final round of coupling, the resin was washed with DMF, DCM, MeOH, and hexane, respectively. Cleavage was performed using a reagent mixture of 2.5% H₂O, 2.5% triisopropylsilane, and 95.0% trifluoroacetic acid for 30 min. The crude product was obtained after precipitation in ether. The crude product of chiral N-pep was directly used for further conjugation. Synthesis of chiral peptide-drug conjugates (N-pepABS) Chial N-pepABS was prepared by conjugating 4-(2-aminoethyl)-benzenesulfonamide on the side chain of C-terminal Lys (Supplementary Fig. 38). First, isothiocyanate-benzensulfonamide (compound 2) was prepared by reacting 4-(2-aminoethyl)-benzenesulfonamide with TCDI. In detail, 3 mmol of ABS was first dissolved in 2 mL DMF. 1.5 mL triethylamine was added to adjust the pH to weakly alkaline (pH about 8.5). After which, 2 mmol of TCDI was fully dissolved in 2 mL DMF in an oil bath at 50 °C, and added dropwise to the above solution. The solution was continuously stirred at room temperature for 30 min monitored by thin layer chromatography (TLC). 10 mL ddH2O was added and stirred for another 15 min to quench the reaction. The product was first extracted with ethyl acetate and then purified by silica gel chromatography (ethyl acetate–hexane) with a yield of around 40%. Purified product was analyzed by 1H NMR, 13C NMR, and mass spectrum (Supplementary Figs. 39–41). Then, Chiral N-pep (0.3 mmol) was dissolved in the mixture of methanol (6 mL) and ddH2O (1 mL), followed by adding 1 M NaOH to adjust the pH value to around 8.5. After which, compound 2 (0.35 mmol) was dissolved in methanol (1 mL) and added to the solution. Additional NaOH was added to maintain the pH value. The mixture was stirred continuously overnight at room temperature, monitored by TLC, and then purified by reverse-phase HPLC to obtain chiral peptide-ABS conjugates, yielding approximately 50%. The final products of chiral N-pepABS were analyzed by 1H NMR, HPLC, CD, and mass spectrometry (Supplementary Figs.1–5 and 42). The molecular structures of the peptides were confirmed by NMR, utilizing d6-DMSO as the solvent, and theoretical prediction for analysis. Synthesis of nitrobenzoxadiazole labeled peptide-ABS conjugates (NBD-DLD /NBD-DDD) 4-Chloro-7-nitrobenzofurazan (NBD-Cl) was conjugated to FDFLKD or FDFDKD via standard SPPS. The crude product (~0.1 mmol) was further dissolved in 4 mL of methanol. The pH value was adjusted to around 9 with 0.1 M NaOH. 0.13 mmol of compound 2 was dissolved in 3 mL of methanol, which was added dropwise to the above solution (Supplementary Fig. 43). The reaction was stirred overnight, monitored by TLC, and purified with RP-HPLC, yielding approximately 50%. The final product was analyzed by 1H NMR and mass spectrum (Supplementary Figs. 44–49). Synthesis of FITC-labeled peptide-ABS conjugates (FITC-DLD/FITC-DDD) Fmoc-β-Ala-FDFLKD or Fmoc-β-Ala-FDFDKD were first synthesized by standard SPPS, followed by conjugation with compound 2, similar to the synthesis of DLD/DDD. A white crude product was precipitated by adding ddH2O and HCl. Then the crude product (~0.05 mmol) was dissolved in MeOH, followed by the addition of 20% piperidine for 30 min to remove Fmoc protection. The solvent was then removed under reduced pressure, followed by the addition of ether to obtain deprotected peptides. Finally, the crude product was reacted with fluorescein isothiocyanate (isomer I, 0.05 mmol) in MeOH/DIEA, and purified by RP-HPLC (Supplementary Fig. 50), yielding approximately 15%. The final product was analyzed by 1H NMR and mass spectrum (Supplementary Figs. 51–56). LC and MS characterization Purified product was analyzed by NMR, MS, or HPLC. Samples synthesized in each batch (n > 3) were fully dissolved in MeOH before analysis. For compound 2, both 1H NMR and 13C NMR were performed on AVANCE III HD 400 (Bruker). HRMS of compound 2 was obtained by Q-Exactive (Thermo Fisher) without a column, employing a flow rate of 0.5 mL/min and a mobile phase composed of pure acetonitrile (as compound 2 is susceptible to acids). For chiral PDCs and fluorophore-labeled PDCs, MS were obtained by LCMS-8040 (Shimadzu) or Q-Exactive, without column, employing a flow rate of 0.3 mL/min and a mobile phase composed of acetonitrile (0.1% TFA) and deionized water (0.1% TFA). HPLC profile of Chiral PDCs was obtained on LC-20AT (Shimadzu) equipped with a C18 column (GL Sciences, INERTSIL ODS-3 5 μm 4.6*250 mm). HPLC profile of fluorophore-labeled PDCs was given on Waters 1525–2489 equipped with a C18 column (XBridge Prep C18 5 μm OBD 19*250 mm). HPLC traces were given in the Supplementary Information. General procedure for sample preparation A 10 mM stock solution was prepared by dissolving chiral peptide-ABS conjugates in 100 mM NaOH to make a clear solution. The salinity and the pH were adjusted with 1 M NaCl and 0.1 M HCl, respectively. ddH2O was added to dilute the sample to a final concentration of 1 mM. The pH value was monitored by special indicator paper (pH 5.5–9.0). For the evaluation of alkaline metal ions. A 100 mM stock solution was prepared by dissolving chiral peptide-ABS conjugates in DMSO, which were dispersed into different buffer solutions containing alkaline metal chloride and 10 mM Tri-HCl (pH 7.5) to a final concentration of 1 mM. Microscopy for phase behavior and droplet fusion Chiral PDCs were prepared at a final concentration of 1 mM containing 150 mM NaCl at pH 7.5. 50 μL samples were added to a 96-well plate for imaging. Phase separation was observed for DLD and LDL. Videos of droplet fusion events were captured at 1 Hz. SHG detection A sample of DLD droplets was excited by a femtosecond laser (1037 nm, 180 fs, 0.2 mJ, 1 kHz). The emitted SHG signal was captured with a spectrometer behind a near-infrared filter. Fluorescence recovery after photobleaching (FRAP) FRAP analysis was performed on a confocal microscope (Olympus FV4000). Briefly, 1 mM DLD containing 150 mM NaCl and 400 nM Cy5 was prepared at pH 7.5. 0.1 M HCl or 20 mg/mL BSA was added to induce liquid-gel transition (final pH at 6.5 or final BSA concentration at 2 mg/mL). Droplets were bleached with a 488 nm laser (2.5% power, 910 ms). Measurement of monomer concentration Chiral peptide-ABS conjugates were dissolved in 0.1 M NaOH and diluted to various concentrations. The UV-Vis spectrums were recorded to construct a standard curve using absorbance at 290 nm (Supplementary Fig. 57). Samples of various salinity were prepared as described above to a final volume of 300 μL, followed by centrifugation at 21,000 × g for 5 min. 100 μL of supernatant was collected, diluted with 100 μL of 0.1 M NaOH for UV-Vis spectrum measurement. The monomer concentration was calculated based on a constructed standard curve. Thioflavin T (ThT) fluorescence assay A stock solution of ThT was first prepared in ethanol (3 mM), which was added to samples of various treatments to a final concentration of 30 μM. The ThT fluorescence was monitored by a PerkinElmer Enspire multimode plate reader (λex = 440 nm/λem = 480 nm). For kinetic trace, ThT fluorescence curves were fitted with a sigmoidal function:\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$F={F}_{0}+\frac{1}{1+\exp [{r}_{\max }\left({\tau }_{1/2}-t\right)]}$$\end{document}F=F0+11+exp[rmaxτ1/2−t]where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\tau }_{1/2}$$\end{document}τ1/2 is the time for half completion of aggregation and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${r}_{\max }$$\end{document}rmax is the maximum growth rate. TEM sample preparation Samples were dropped on a carbon-coated grid (230-mesh copper grids that had been coated with continuous thick carbon film) and then negative-stained with 2% uranyl acetate for 1 min, after which the copper grids were dried in a desiccator and observed using Tecnai G2 20 S-TWIN electron microscope at 200 KV. XPS sample preparation LLPS of DLD was induced at 150 mM NaCl, pH 7.5. The coacervate phase was enriched by centrifugation at 21,000 × g for 5 min. The supernatant was removed, followed by adding ddH2O to wash the interface of the water/coacervate phase. The coacervate phase was washed twice and lyophilized for XPS analysis. DFT simulation All quantum chemical (QM) calculations were conducted to gain a better understanding of the behavior of self-assembly by exploring the structures and interactions of the peptide-assembled nanostructures and M+ (M = Li, Na, K, Rb, and Cs). The peptide conformations from the reported first-principles calculations were used for QM calculations53. Peptide-M+ molecular structures were optimized with the Gaussian 16 program at the B3LYP + D3(BJ)/def2TZVP level of theory53–55. Dispersion correction was accounted for using Grimme-D3(BJ)56,57. The peptide conformational sampling step was carried out CREST58–60. Conformation of global minima not accepted to investigate the binding with Na+, as the conformation is folded, and exhibited lower binding energy than the discussed binding mode (Supplementary Fig. 58). Circular dichroism (CD) spectrum All samples were prepared for CD analysis, including dilution in ddH2O (0.01 wt%) and in methanol (0.01 wt%) at different pHs. Then, we measured CD spectra by using JASCO J-1500 in a 1 mm quartz cuvette with a bandwidth set to 2 nm and a scanning speed set to 200 nm/min. Cell lines and cell culture MDA-MB-231 (CL-0150) and 4T1 (CL-0007) cell lines were purchased from Procell with STR authentication reports. MDA-MB-231 cells were cultured with Dulbecco’s Modified Eagle Media (DMEM) containing 10% fetal bovine serum (FBS) and 1% penicillin-streptomycin. 4T1 cells were cultured with RPMI-1640 medium containing 10% FBS and 1% penicillin streptomycin. All cells were maintained at 37 °C in a 5% CO2 incubator. For hypoxia experiments, before being treated with short peptide–constructed CA inhibitors, cells were precultured in the hypoxic incubator for about 12 h at 37 °C in a hypoxic incubator with 1% O2 and 5% CO2. Cell viability assay About 1 × 104 cells were seeded into a 96-well plate (100 μL/well), and cultured overnight in a 5% CO2 incubator or in a hypoxic incubator at 37 °C. Then the cells were treated with DLD or DDD prepared under different salinities. After 24 h, the cell viability was measured with a CCK-8 kit. Fluorescent Imaging About 1 × 103 cells were seeded into an 8-chambered coverglass system and cultured overnight in a hypoxic incubator at 37 °C. Then the cells were treated with DLD containing 4% NBD-DLD or DDD containing 4% NBD-DDD. After different time points, cells were washed with fresh culture medium, fixed with 4% paraformaldehyde, stained with DAPI, and observed with a confocal microscope. Flow cytometry Around 5 × 104 MDA-MB-231 cells were seeded in a 12-well plate and cultured overnight in a hypoxic incubator at 37 °C. According to literature, cells were pretreated by Methyl-β-cyclodextrin (2.5 mM) and Amiloride (20 μM) or low temperature for 1 hour. Then 200 μM DLD (containing 1% FITC-DLD) was added. 4 h later, cells were harvested for flow cytometry. To evaluate the role of hypoxia, around 5 × 104 MDA-MB-231 cells were seeded in a 12-well plate, with or without hypoxia treatment overnight. Then 200 μM DLD (containing 1% FITC-DLD) was added. 4 h later, cells were harvested for flow cytometry. Gating strategy is given in the Supplementary Information (Supplementary Fig. 59) Fluorescent correlation spectrum (FCS) characterization FCS experiments were performed using a commercial bench-top fluorescence correlation spectrometer (CorTectorTM SX200, LightEdge Technologies Limited, China) equipped with a 60 × 1.2NA UPLANSAPO water immersion objective (Olympus, Japan) and two (488 nm, 638 nm) continuous-wave solid-state lasers (Pavilion Integration Corporation, China). The FCS instrument is based on a confocal optical pathway, using a multimode optical fiber with an inner core diameter of 50 μm as the confocal pinhole. The 488 nm laser was used to excite the dyes, detected by an optical fiber-coupled single-photon counting avalanche photodiode detector (SPCM-800-14-FC, Excelitas Technologies Corporation, USA). For each experiment session, the confocal pinhole was first automatically aligned using a 10 nM ATTO655 carboxylic acid (ATTO-TEC GmbH, Germany) solution. FCS data of the ATTO655 calibration sample was collected and analyzed using the Correlation Acquisition and Correlation Analysis software (LightEdge Technologies Limited, China), respectively. Using the known diffusion of 4.26 × 106 cm2/s of the ATTO655 dye, the structure factor (S) and confocal volume (V) of the FCS instrument in its current state were determined, which were used for subsequent sample data analysis. A sensor in the sample chamber of the instrument recorded the experimental temperature, which was then used to automatically correct the temperature effect on molecular diffusion during data analysis. For the FCS experiments, 500 nM FITC-DLD was added to 1 mM DLD in 5 mM Tris buffer, before the addition of NaCl to a final concentration of 150 mM to induce LLPS. Then, an equal volume of FBS was added. Diluted samples were used for FCS analysis. Meanwhile, FCS of 500 nM FITC-DLD alone in NaCl/Tris buffer and a diluted FBS solution was also obtained. Animal experiments An animal model was built up and analyzed, depending on the Institutional Animal Use and Care Committee at the Chinese Academy of Sciences. Mice were housed at 20–22 °C with a 12 h light/dark cycle and at 30–70% humidity. The maximal tumor size permitted by the ethics committee is 1500 mm3, and the maximal tumor size was not exceeded in this study. About 6-week-old female BALB/c nude mice were injected with MDA-MB-231 cells subcutaneously. When the tumors were detectable, the mice were randomly divided into 3 groups (n = 5): control group (Control), N-pep (Peptide alone) group, and N-pepABS (PDC) group. The tumor volume was calculated according to the following equation: volume = 1/2(tumor length) × (tumor width)2. Then, N-pep (1 mM, 100 μL) or N-pepABS (1 mM, 100 μL) was intravenously injected into the mice every 3 days. After being treated for several days, the mice were sacrificed, while each organ or tumor was harvested for further analysis. To compare anti-tumoral effects between DDD and DLD, about 6-week-old female BALB/c nude mice were injected with 4T1 cells subcutaneously. When the tumors were around 50 mm3, the mice were randomly divided into 3 groups (n = 5): control group (Control), DDD-treated (DDD) group, and DLD-treated (DLD) group. Both DDD (1 mM, 200 μL) and DLD (2 mM, 100 μL) were intravenously injected every 2 days. The tumor volume and body weight were monitored every 2 days. When the tumor volume of one mouse reached 1500 mm3, the mice were sacrificed to harvest the tumor and major organs for analysis. For the investigation of biodistribution, 4T1 tumor-bearing mice were injected with DDD or DLD containing 5% of FITC-DDD or FITC-DLD. At a certain time point post-injection, mice were sacrificed, while each organ or tumor was harvested for imaging and further analysis. Statistics and reproducibility The data are expressed as mean ± SD from at least three repeated experiments. Statistical differences between the two groups were assessed by a two-tailed unpaired t-test. The bar graph was represented as means ± SD, and *P < 0.05 was considered statistically significant. Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Supplementary information Supplementary Information Reporting Summary Transparent Peer Review file Supplementary Information Reporting Summary Transparent Peer Review file Source data Source Data Source Data