Efficient intracellular delivery of nucleic acid (NA) fragments and derivatives is important for fundamental research and therapeutic strategies involving the use of antisense oligonucleotides, synthetic small interfering RNA, guide RNA, etc., for mRNA targeting or gene editing (Sabatos-Peyton et al., 2010; Vickery et al., 2011; Bartolucci et al., 2022). Liposomal, polymeric and micellar nanoparticles are commonly used as delivery vehicles for mRNA or antisense oligonucleotides (Lorscheider et al., 2021; Pacheco et al., 2023). Their limitations include the lack of biodegradation, low bioavailability toxicity, etc. (Chong et al., 2014; Gupta et al., 2019; Kumawat et al., 2019; Lu et al., 2020; Wang et al., 2022).

Coacervate-based delivery systems are a promising modern alternative to classical (liposomal or polymeric) nanoparticles. NA incorporation occurs throughout the coacervate volume due to multiple weak electrostatic interactions and is characterized by high efficiency. Furthermore, coacervates effectively protect therapeutic NAs from biodegradation (Sun et al., 2022). The main advantages of coacervates over classical nanoparticles include biocompatibility, ease of cell penetration, and low cytotoxicity (Jo et al., 2019; Xiao et al., 2020). In addition to delivering NAs or low-molecular-weight agents to tumor cells (Liu et al., 2022), coacervates and similar hydrogel-type systems have been used to deliver protein growth factors to damaged tissues (Kim et al., 2018; Wu et al., 2018). These applications demonstrate the potential of coacervates in regenerative medicine, oncotherapy, and allergy treatment.

Of the peptide-based coacervates being considered as delivery systems, HBpep coacervates are the most thoroughly characterized (Lim et al., 2018; 2020; Shebanova et al., 2022; Sun et al., 2022; 2025; Liu et al., 2023; Pramanik et al., 2025). The HBpep sequence (GHGVY GHGVY GHGPY GHGPY GHGLY W), derived from the squid beak protein HBP (Tan et al., 2015), shows high propensity for coacervation due to multiple His and Gly residues (Blocher and Perry, 2017). The coacervates successfully delivered therapeutic agents into cells but remained stable for several days, resulting in inefficient drug release (Lim et al., 2018). To solve this problem, Sun et al. modified the peptide design to make it redox-sensitive (Sun et al., 2022). Specifically, they inserted a single Lys residue into HBpep, which shifted the optimal pH for coacervation from 7.5 to 9.0, and masked this Lys residue with a redox-sensitive (disulfide bond-containing) “self-destructive” moiety. The resulting HBpep variant (HBpep-SR) formed coacervates that readily incorporated guest macromolecules at pH 6.5 and released them after crossing the cellular membrane in response to intracellular glutathione. The only drawback of this approach is the complexity of HBpep-SR synthesis.

Here, we report a different approach to fine-tuning coacervates for efficient drug release. To facilitate intracellular disassembly, we introduced an ATP-sensitive module into the HBpep sequence. This modification is compatible with the standard peptide synthesis and thus might be more convenient than HBpep-SR. We report the selection of an optimal ATP-sensitive module and verification of its effect on the coacervate packing and unpacking.

We aimed to obtain HBpep-based peptides that assemble into ATP-sensitive coacervates for ATP-driven intracellular disassembly. The ability of ATP to solubilize peptides and proteins is due to its amphiphilic nature. The combination of a charged fragment (phosphate) and an aromatic fragment (nucleobase) enables ATP to disrupt typical coacervate contacts, including π-π/π-cation interactions between aromatic residues and the cationic amino acid residues, as well as ionic bonds, van der Waals interactions, and hydrogen bonds (Nott et al., 2015; Riback et al., 2017; Vernon et al., 2018; Chiu et al., 2020).

We searched for short ATP-binding motifs to introduce into the HBpep sequence and selected consensus motifs that are overrepresented in human kinases (Xiao et al., 2013) and fragments of viral proteins (Mujeeb et al., 1994; Titolo et al., 1999; Scholz, 2003). Motif 1 (the P-loop sequence GxxxxGK) was identified by screening biotin-labeled peptides from all human ATP-binding proteins, and motif 2 (HRDxKxxN) was derived exclusively from kinases (Xiao et al., 2013). The viral motifs include the ATP-binding sites of the terminase subunit pUL56 of the human cytomegalovirus (Scholz, 2003) (motifs 3, 4, and 5) and the helicase of the human papillomavirus type 11 E1 (Titolo et al., 1999) (motif 6). Table 1 shows the sequences of the resulting peptides with the ATP-binding modules (HBpep-ATP-1–6), as well as the control peptide lacking the ATP module (HBpep-contr). Initially, we added each ATP-sensitive module to the C-terminus of the peptide (HBpep-ATP-1–6). After the first round of assays, we selected a leader, and added its analog with the ATP-module shifted to the N-terminus (HBpep-ATP-1Nt) to the peptide set.

To facilitate imaging of the coacervates, the peptides were labeled at their N-termini with a fluorescein isothiocyanate (FITC) residue, which fluoresces in the green range of the spectrum. The labeled peptides were then mixed 1:99 with the unlabeled peptides. Random RNA was added to stimulate coacervation. In line with previous reports for HBpep (Harrison and Shorter, 2017), peptide coacervation was inefficient in weakly acidic aqueous solutions and in the absence of RNA (Figure 1A). After adding RNA in a pH 7.4 buffer, peptides HBpep-ATP-1(Nt), HBpep-ATP-3, and HBpep-ATP-6 (Supplementary Figures S1, S2) exhibited good coacervation, similar to HBpep-ATP-contr (Supplementary Figures S3). The resulting coacervates appeared as brightly colored, mobile, micrometer-sized droplets that partially wetted the glass surface (Figure 1A). Their circularity was ≥0.8. The coacervates were stable in ATP-free solutions for at least 24 h. HBpep-ATP-2 separated with RNA, but some of the resulting droplets had irregular shapes (circularity ≤0.6), which suggests gelation or solidification. In the cases of HBpep-ATP-4 and HBpep-ATP-5, phase separation was minor (Supplementary Figure S4). We conclude that ATP-sensitive motifs 2, 4, and 5, but not motifs 1, 3, and 6, hamper droplet formation.

The liquid state of the observed droplets was confirmed using fluorescence recovery after photobleaching (FRAP) (Figures 1B,C). A fragment of a droplet from the FITC-labeled HBpep-ATP-1-containing sample was bleached, and the fluorescence recovery was monitored for 30 min. Representative images are shown in Figure 1B. The recovery reached ≈65% (Figure 1C) and was moderately slow, with an apparent tau_1/2 of 190 ± 30 s, which is consistent with the dense liquid state. Therefore, the observed droplets are indeed coacervates rather than solid aggregates.

To verify the ability of coacervates to incorporate nucleic acid therapeutics, we used a G-quadruplex-forming fragment of human telomeric repeats (GGGTTA)₃GGG (22AG) as a model ODN. Quadruplexes have recently emerged as a promising new type of anticancer drug candidates. The first FDA-approved therapeutic of this type (World Health Organization, 2024), Imetelstat (RYTELO™), is a lipid-conjugated oligonucleotide (Carloni et al., 2021; Keam, 2024) that inhibits telomerase and can be regarded as a 22AG mimic.

The SIMA-labeled model ODN (SIMA-22AG), that fluoresces in the red channel, was added to a mixture of unlabeled and FITC-labeled peptides at pH 7.4. This had no significant effect on the morphology of the coacervates visualized in the green channel (Supplementary Figures S1-S3). The colocalization of coacervates in the red and green channels (Figure 1D) confirms the ODN incorporation. A few ODN-free droplets were observed, and the colocalization seemed to be incomplete for small droplets, likely due to their rapid movement revealed by visual screening. Nevertheless, the apparent colocalization was within the range of 75%–90% for all HBpep variants, and the ODN loading efficiency (60% ± 20% for HBpep-ATP-1, 40% ± 3% for HBpep-ATP-3, 15% ± 2% for HBpep-ATP-6, 38% ± 9% for HBpep-ATP-1Nt, and 40% ± 10% for HBpep-ATP-contr) varied in accordance with the total coacervate projection area (Figure 1E).

Next, we tested the efficiency of ATP-driven coacervate unpacking in vitro at ATP concentrations ranging from 2.5 to 20 mM. Coacervates of all peptides, except HBpep-contr, with random RNA and SIMA-labeled ODN showed clear, dose-dependent disassembly (Supplementary Figure S1-S3). It was evident from the analysis of the total coacervate projection area (Figure 1E). In contrast, the ATP-dependence of HBpep-contr coacervates was irregular and the least pronounced (Figure 1E; Supplementary Figure S3). HBpep-ATP-3 coacervates also exhibited an ATP-dependent decrease in average coacervate size (Figure 1F; Supplementary Figure S2).

Before testing the coacervates as intracellular delivery vehicles, we evaluated their cytotoxicity. Figure 2A summarizes their effects on the viability of HEK293 cells. Most of the coacervates were nontoxic at concentrations up to 1 mg/mL, with CC₅₀ values exceeding 1.25 mg/mL in all cases.

Next, we tested the coacervate-based delivery of the model oligonucleotide 22AG to HEK293 cells. We monitored the internalization of FITC-labeled HBpep-ATP-based coacervates (green) and SIMA-labeled 22AG (red) using fluorescence microscopy in real time (Figure 2B; Supplementary Figures S5-S7). The resulting kinetic curves are shown in Figure 2C. HBpep-ATP-3- and HBpep-ATP-6-based coacervates either failed to enter cells or were rather unstable in the biological media (despite the reasonable stability in PBS) and disassembled prior to cellular uptake. HBpep-contr-based coacervates entered cells but remained stable after the uptake for 24 h and showed inefficient 22AG release (Supplementary Figure S5). HBpep-ATP-1-based coacervates entered cells within 4 h and released SIMA-22AG into the cytoplasm within 20 h (Supplementary Figure S6). Thus, HBpep-ATP-1 was the most promising peptide from the initial set with the ATP module at the C-terminus. Because the position of the module can affect the peptide’s physical properties and performance in delivery assays, we obtained an HBpep-ATP-1 analog, HBpep-ATP-1Nt, with the same kinase-derived ATP module at the N-terminus, and repeated all the experiments. The performance of HBpep-ATP-1Nt in vitro was mostly similar to that of HBpep-ATP-1 (Figures 1D,E; Supplementary Figure S1), while in HEK-293 cells it slightly outperformed HBpep-ATP-1 in oligonucleotide delivery assays (Figures 2B,C; Supplementary Figures S6, S7).

Finally, we examined the potential use of the new coacervates in delivering long therapeutic nucleic acids, such as mRNA or DNA, which are being used more frequently as vaccine components. To do so, we used a model mRNA that encodes the Firefly luciferase reporter protein (FLuc), a model plasmid that contains the same reporter gene (pCMV-Fluc), and a plasmid that encodes green fluorescent protein (pEGFP-N1). We assessed the efficiency of mRNA/pCMV-Fluc delivery to HEK293 cells in HBpep-ATP-1 coacervates using standard luciferase assays, and the efficiency of pEGFP-N1 delivery in HBpep-ATP-1, HBpep-ATP-1Nt, or HBpep-contr coacervates using fluorescence microscopy imaging 1 day after transfection (Figure 2D). For the control assays, we used standard transfection reagents Messenger Max and Lipofectamine 3,000 for mRNA and DNA, respectively. HBpep-ATP-1 substantially underperformed to the known transfectants in Fluc assays and was inferior to HBpep-contr in pEGFP delivery assays. HBpep-ATP-1Nt slightly outperformed HBpep-ATP-1 in the pEGFP assays but was still substantially inferior to HBpep-contr.

To gain insight into possible reasons for the poor performance of HBpep-ATP-1 in the plasmid delivery assay compared to the ODN delivery assay, we performed additional FRAP experiments with 22AG-loaded and pEGFP-loaded coacervates (Supplementary Figure S8) and compared the results to those obtained for the cargo-free coacervates (Figure 1B). ODN 22AG caused no substantial changes in the recovery kinetics of FITC-HBpep-ATP-1 (tau_1/2 = 220 ± 20 s), whereas pEGFP slowed it down significantly (tau_1/2 = 440 ± 30 s). The latter result suggests reduced diffusion within coacervates, which suggests reduced DNA release and may partly explain reduced expression.

Thus, HBpep-contr is suitable for the intracellular delivery of the lengthy DNA/RNA, probably because such a cargo can be released in an ATP-independent manner. In contrast, the ATP module-containing peptides HBpep-ATP-1 and HBpep-ATP-1Nt are tuned for short oligonucleotides. Lengthy nucleic acids can form multiple transient interactions with the ATP modules, yielding coacervates of increased density, which may account for inefficient unpacking. Thus, future studies of the coacervates could include determination of the threshold nucleic acid size and the optimization of the peptide:nucleic acid ratio. Additionally, the mixed-sequence random RNA could be substituted for a more appropriate homogeneous biopolymer.

To summarize, we have demonstrated that HBpep-ATPs form coacervates that incorporate different NAs. These cargo-loaded coacervates are engulfed by HEK293 cells and disassemble, supposedly due to the relatively high intracellular ATP levels (0.2–15.8 mM (Greiner and Glonek, 2021)), releasing the cargo into the cytosol. The precise uptake mechanism of such peptide coacervates awaits clarification. Based on the previous studies of other coacervates, lipid-raft-mediated endocytosis appears to be the likely one (Vedekhina et al., 2025). The coacervates based on HBpep variants with the human kinase-derived ATP module were superior to unmodified HBpep in assays with the model ODN 22AG, making them promising platforms for intracellular delivery of short oligonucleotide therapeutics.