From LNPs to hybrid nanocarriers: development, challenges and redesign of non-viral gene delivery Authors: Wenlin Wu, Zhigong Wei, Xingchen Peng Published: 2026 ABSTRACT AbstractGene therapy has become a cornerstone of precision medicine, offering potential solutions for a wide spectrum of disorders at the molecular origin. However, the clinical performance of gene therapeutics is critically determined by the efficiency and safety of delivery vectors. Though viral vectors are widely used in gene therapy because of their high delivery efficiency, the safety concern remains a major limitation. In contrast, non-viral vectors have attracted increasing attention owing to their favorable safety profiles, structural tunability, larger cargo capacity, and scalable production. In this review, we examine five major classes of non-viral gene delivery vectors, including Lipid Nanoparticles (LNP), polyplexes, lipoplexes, and exosomes, together with their corresponding hybrid systems. For each platform, we summarize structural features, delivery mechanisms, therapeutic applications, and translational challenges, and further discuss how hybrid architectures integrate complementary properties from different vectors to overcome limitations inherent to single-component systems. By organizing these delivery strategies within a unified structure–function–application framework, this review provides an integrated perspective on non-viral vector design and highlights hybrid systems as an increasingly important direction for the optimization of clinically relevant gene delivery platforms.Graphical abstract Keywords: Non-viral vectors, Lipid Nanoparticles (LNP), Polyplexes, Lipoplexes, Exosomes, Hybrid vectors Introduction Gene therapy refers to the introduction of nucleic acids into cells to correct or modulate defective functions. By targeting the molecular basis of diseases, it offers the prospect of durable or even curative outcomes. Over the past decades, this strategy has evolved into a clinical reality, providing more options for diseases that used to be hard to cure. Gene therapy is increasingly applied to complex conditions, including rare monogenic disorders, cardiovascular, neurodegenerative disorders, and cancer, positioning itself as a central component of modern precision medicine. However, when we deliver genetic material into the body, some problems and barriers exist: endonucleases in serum, mononuclear phagocyte system, cellular barriers, the uncertain half-life of DNA [1] and systemic barriers, such as the blood-brain barriers (BBB) which result in low gene delivery efficiency to specific sites (Fig. 1). Against this backdrop, the choice of delivery vector, viral or non-viral, has become a critical determinant of successful delivery. (a) receptor-mediated endocytosis, (b) early/late endosome formation, (c) endosomal escape, and (d) post-escape cargo routing. Viral vectors remain the most widely used vehicles for gene delivery owing to their high transduction efficiency. However, their clinical application is hampered by safety concerns, including the risks of insertional mutagenesis and oncogenesis, as well as inherent limitations such as restricted packaging capacity and complex large-scale manufacturing [2, 3]. In contrast, non-viral vectors offer several advantages, including lower oncogenic risk, larger loading capacity, and more straightforward production. These unique features have positioned non-viral platforms as increasingly important candidates in the field of gene therapy. With more and more non-viral vectors emerging, they play crucial roles in gene therapy field. In this review, we highlight several representative non-viral approaches: Lipid nanoparticles (LNP), polyplexes, lipoplexes, exosomes and their hybrid system. We first outline the mechanisms, applications, challenges, and emerging frontiers of each strategy. Then we compare their respective strengths and limitations, followed by a synthesis of insights from ongoing clinical trials. Finally, we highlight the emerging trend toward hybrid non-viral delivery systems, integrating lipid, polymer, and exosomes, as the next generation vectors for achieving programmable, safe, and clinically scalable gene therapeutics. By combining mechanisms with clinical translation across non-viral delivery systems, this review aims to construct a framework that bridges the gap between design and vector function, thereby guiding further optimization for clinical gene therapy applications. Lipid nanoparticles (LNP) Lipid Nanoparticles, the leading vector in non-viral systems, originated from Liposomes. In 1990s, Müller and Lucks in Germany brought up Solid Lipid Nanoparticles (SLNs) which replace liquid lipid with a solid one. The change elevates the stability of Liposomes but also largely limits the capacity of cargo [4]. Besides, SLNs are prone to crystallization during storage, leading to drug leakage. Then, Nanostructured Lipid Carriers (NLCs) emerged. They were refined by replacing some of the solid lipid components in SLNs with liquid lipids, leading to an increased capacity for drug incorporation [4]. The applications of NLCs are mainly reflected in the delivery of various targeted drugs, such as lung cancer, breast cancer, gastric cancer and lymphoma. Drug-co-loaded NLCs demonstrate enhanced transfection efficiency and effectiveness in targeting cancer cells, while reducing systemic toxicity. Various drug-loaded NLC formulations, such as Paclitaxel-doxorubicin co-loaded NLCs for lung cancer, β-lapachone and doxorubicin co-loaded NLCs [5], thymoquinone loaded NLCs [6] for breast cancer. Although NLCs have shown certain efficacy in drug delivery, their low encapsulation efficiency and poor cellular internalization limit their applicability for nucleic acid delivery. Consequently, LNPs have emerged as a promising alternative for nucleic acid delivery. The basic components of LNP are: Ionizable lipid, helper lipid, cholesterol and PEG-lipid (Fig. 2). In contrast, formulations containing permanently cationic lipids such as DOTMA or DOTAP are generally classified as cationic liposomes rather than LNPs. Cationic liposomes exhibit high nucleic acid complexation efficiency due to strong electrostatic interactions. However, their positive charge promotes non-specific interactions with plasma proteins and cellular membranes, which may trigger unexpected immune responses. These limitations were overcome with the advent of the ionizable lipid. It is characterized by a neutral charge in plasma, which enhances tolerability and reduces cytotoxicity, whereas under acidic conditions such as endosomes (pH 5–6), it acquires a positive charge, thereby promoting efficient fusion with the endosomal membrane and facilitating nucleic acid release, which is called Endosomal Escape. The pH-titratable properties made ionizable lipids widely used in nucleic acid therapy and many vaccines as they make great improvement in overall safety and better transduction efficacy [7]. Chemical constitutions of Lipid Nanoparticles (LNP): LNP is constituted of (a) cholesterol, (b) helper lipids (DSPC and DOPE), (c) PEG lipids (PEG-2000-C-DMG and ALC-0159) and (d) ionizable lipids (DLin-MC3-DMA, SM102 and ALC-0315) Clinical use of LNP Genetic diseases The first application of LNP is Patisiran (Onpattro®) which uses DLin-MC3-DMA to deliver siRNA to hepatic cells for the treatment of Hereditary Transthyretin-Mediated Amyloidosis. It is the first FDA-approved siRNA-LNP therapeutic delivered by LNP [8]. In the pivotal APOLLO trial, patisiran reduced serum TTR by approximately 80%. and produced a least-squares mean difference of 34.0 points relative to placebo in mNIS + 7 at 18 months, although the transient nature of RNAi requires lifelong intravenous infusion every three weeks [9]. This clinical success is enabled in part by DLin-MC3-DMA, the core ionizable lipid widely used in LNP, which exhibits excellent safety and tolerability. However, its serum half-life of approximately 70 h and lack biodegradability may lead to lipid accumulation in the body in indications requiring frequent dosing, thereby increasing the long-term metabolic burden and limiting its potential applications [8]. To settle the safety problem, L319 lipid is created. Its ester-containing hydrophobic tails confer rapid enzymatic degradation in vivo, significantly reducing long-term tissue accumulation. Collectively, these features endow L319 with superior pharmacokinetic and safety profiles, making it a promising platform lipid for vaccines, protein replacement therapies, and other mRNA-based interventions [10]. Beyond chronic RNAi-based silencing, LNP-delivered gene editing has emerged as a strategy for durable, single-dose correction of monogenic disease. NTLA-2001 (Intellia; NCT04601051), an LNP co-formulating Cas9 mRNA with a TTR-targeting guide RNA, achieved a mean reduction in serum TTR of approximately 87% at 28 days after a single intravenous infusion in patients with hereditary transthyretin amyloidosis, with knockdown sustained for at least 12 months [11]. VERVE-101 and VERVE-102, which deliver an LNP-encapsulated base editor targeting PCSK9, similarly demonstrated dose-dependent and durable LDL-C reduction after a single infusion in patients with heterozygous familial hypercholesterolemia (NCT05398029). These programs illustrate a fundamental shift in the LNP dosing paradigm: from chronic repeat administration (patisiran, every three weeks) to single-dose curative editing (NTLA-2001, VERVE-101). Meanwhile, the mRNA-based protein replacement (e.g., Moderna mRNA-3704 for methylmalonic acidemia) occupies an intermediate position requiring repeat dosing every 1–3 weeks due to the transient nature of mRNA expression. Now, ionizable lipid becomes the main force of LNP. Ongoing refinements to ionizable lipids are being pursued. Recent studies have increased the number of hydrogen-bond donors to minimize complement activation and non-specific interactions with serum proteins, and optimized the lipid-to-mRNA ratio to maintain a nearly neutral surface charge, thereby mitigating undesired immune responses and systemic side effects [12]. Another research by Hashiba found the piperidine-based lipids improve the long term storage stability of LNPs at refrigeration temperature as a liquid formulation [13]. The amelioration of ionizable lipids has led to their increasingly widespread application in various fields. H5N1 influenza vaccines formulated with LNPssPalmO increase protective immunity and fewer adverse effects than conventional SM-102-based LNPs [14]. A new kind of CrLNPs which deliver CRISPR/Cas9 ribonucleoprotein shows high efficiency in vivo and in vitro [15]. This approach effectively mitigates the low target specificity of CRISPR-Cas9 while maintaining relatively low toxicity. Besides, LNPs function as versatile carriers, delivering nucleic acids or small molecule drugs to various cells in the body to achieve the purpose of treating diseases, such as hemophilia A [16], ophthalmic diseases [17], Glioblastoma(GBM) [18] and Lysosomal storage disorders [19]. Taking lysosomal storage disorders as an example, delivery of plasmid DNA via LNPs enables intracellular production of lysosomal enzymes, thereby allowing the proteins to undergo native processing and trafficking pathways. Following expression, these enzymes are folded and modified within the endoplasmic reticulum and Golgi apparatus and can be correctly targeted to lysosomes, typically through mannose-6-phosphate–dependent mechanisms. This intracellular biosynthesis is particularly important, as proper subcellular localization is essential for enzymatic activity and is often difficult to achieve through administration of recombinant proteins alone [20, 21]. Unlike conventional enzyme replacement therapies, which are limited by rapid clearance and poor tissue penetration, LNP allows intracellular production of therapeutic proteins, thereby addressing both enzyme deficiency and biodistribution challenges. Moreover, by delivering DNA, the vector shows sustained expression of functional lysosomal enzymes in target tissues. The capacity of LNPs to encapsulate arbitrary nucleic acid sequences without altering the carrier chemistry has further enabled patient-specific (“N-of-1”) therapeutics, in which the cargo rather than the formulation is individualized. Beyond personalized neoantigen vaccines, recent proof-of-concept work has extended this paradigm to ultra-rare monogenic disease: a custom guide RNA targeting a private pathogenic variant was co-formulated with adenine base editor mRNA into an LNP and administered to an infant with severe carbamoyl phosphate synthetase 1 (CPS1) deficiency within months of diagnosis, producing measurable correction of urea-cycle function after a single infusion [22]. The LNP platform is sequence-agnostic, so only the guide RNA needs to be patient-specific. Regulatory standards and long-term safety monitoring for N-of-1 LNP therapies, however, are still being defined. Pandemic diseases In 2020–2021, COVID-19 emerged worldwide, creating an urgent need for vaccines. And mRNA vaccines stand out because of their stability and safety. In the trials of BNT162b2 of BioNtech, the case split corresponds to 95.0% vaccine efficacy (95% confidence interval [CI], 90.3 to 97.6) [23] and mRNA-1273 of Moderna [24]. Although both the Moderna and BNT162b2 vaccines have shown remarkable efficacy, their potential risks are being progressively elucidated through ongoing research as their widespread use continues. The pandemic experience also exposed several characteristic limitations of LNP-delivered mRNA that have shaped subsequent clinical use. The duration of mRNA-induced humoral immunity is limited: neutralizing antibody titers wane substantially within 4–6 months after the primary two-dose series, necessitating booster administrations to restore protection, particularly against emerging variants. This dosing pattern reflects the intrinsically transient nature of mRNA expression rather than a vector-specific defect, but it has direct implications for the long-term cost, logistics, and patient compliance of LNP-based vaccination programs. Moreover, repeated administration has been associated with the induction of anti-PEG antibodies and rare reactogenicity events including myocarditis (predominantly in young males after mRNA-1273) and complement activation–related pseudoallergy, which is discussed in the challenges of LNP. Building on this clinical foundation, the LNP–mRNA platform has been rapidly extended to other infectious diseases. mRNA-1345 (Moderna), an LNP–mRNA vaccine encoding prefusion-stabilized RSV F glycoprotein, has progressed through Phase 1–3 evaluation in older adults (NCT04528719) and represents the first non-COVID respiratory mRNA vaccine to advance toward approval [25, 26]. LNP–mRNA vaccines against pandemic-potential influenza strains H10N8 and H7N9 (NCT03076385), seasonal influenza, and rabies (NCT03713086) have also entered clinical evaluation [27], while next-generation low-inflammation lipids such as ssPalmO have been shown to elicit protective immunity against H5N1 influenza with reduced reactogenicity compared with conventional SM-102 LNPs in preclinical models [14]. Beyond viral vaccines, TKM-130,803, an LNP–siRNA targeting Ebola virus polymerase, advanced to Phase 2 evaluation during the West African outbreak (PACTR201501000997429), demonstrating that LNP delivery is not restricted to mRNA cargoes in infectious-disease settings [28, 29]. Cancer therapy In the realm of oncology, notable progress in LNP has been made with individualized neoantigen mRNA vaccines. The combination of the personalized cancer vaccine mRNA-4157 (Moderna/Merck) with anti PD-1 therapy is being rigorously evaluated in the adjuvant setting for high-risk melanoma, with a Phase II randomized study (KEYNOTE-942, NCT03897881) reporting a meaningful improvement in recurrence-free survival compared to anti-PD-1 monotherapy [30–32]. KEYNOTE-942 is constituted of SM-102, DSPC, cholesterol, PEG2000-DMG and personalized encoding mRNA. The rational design of these lipid components ensures efficient delivery and intracellular release of mRNA, moreover, enables the encapsulated patient-specific neoantigens to elicit highly targeted and personalized antitumor immune responses. Moreover, ligand-modified LNPs targeting antigen-presenting cells, particularly dendritic cells in lymph nodes, have been shown to improve mRNA vaccine efficacy by promoting antigen presentation and subsequent activation of tumor-specific immune cell responses. Mannosylated LNPs (STLNPs-Man) which targeted Mannose receptor (CD206) achieve four-fold higher DC uptake compared to conventional LNPs after intramuscular injection. Remarkably, these formulations achieve comparable therapeutic efficacy at one-fifth the injection dose of standard LNPs, significantly reducing side effects [33, 34]. An additional immunomodulatory benefit is that the mannose-modified LNPs downregulate CTLA-4 expression on T cells through blockade of the CD206/CD45 axis, enhancing synergy with immune checkpoint blockade [33]. Similarly, dual-targeted LNPs functionalized with DEC-205 antibodies (fused with ApoA1 for simplified lipid binding) simultaneously target both DCs and tumor cells, addressing the problem of immune escape from heterogeneous antigen expression. Co-delivery of antigen-encoding mRNA and TLR agonists via these dLNPs elicited robust CD8⁺ T cell responses and potent antitumor effects without systemic toxicity [35]. Targeted LNPs have demonstrated promising applications in directly modulating immune cells. This targeted strategy not only significantly increases delivery efficiency, allowing dose reduction and minimizing systemic toxicity, but also enables additional immunomodulatory effects, such as checkpoint regulation and synergistic immune activation. Immune regulatory therapies Beyond cancer vaccines, a compelling area of experimental research focuses on leveraging LNPs for local delivery of mRNA encoding cytokines and immune regulatory molecules. This approach aims to remodel the tumor microenvironment and enhance existing immunotherapies. For instance, LNPs have been engineered to deliver mRNA coding for cytokines like IL-12 and GM-CSF, or even regulatory elements such as PD-L1 siRNA. A recent sophisticated example demonstrates the therapeutic potential of the intratumoral delivery of LNP-formulated mRNA encoding a ‘triplet’ of immunomodulators: IL-21, IL-7, and 4-1BB ligand (4-1BBL). This localized delivery strategy was shown to effectively resensitize resistant tumor models to anti-PD-1 therapy, highlighting the potential of LNP as a platform to trigger anti-tumor immunity and overcome resistance [36]. Challenges of LNP Although the use of LNPs has become increasingly mature and clinically established, some problems still appear. The most troubling problem is the excessive hepatic tropism and low endosomal escape efficiency leading to poor extrahepatic delivery and low overall utilization. Hepatic tropism can be explained by the interactions with the reticuloendothelial system (RES) and Kupffer cells, which clear foreign nanoparticles from the bloodstream [37]. Besides, the serum lipoprotein ApoE binds to LNPs, leading to delivery via low-density lipoprotein receptor (LDLR) expressed on hepatocytes [38]. High blood flow of liver, the electric charge and surface-modified structure of LNP [39–41], all of these contribute to the liver targeted accumulation of LNPs, and finally result in hepatic lesion and poor bioavailability. Another bottleneck is the Endosomal Escape. LNPs are internalized by cells via endocytic pathways, including clathrin-mediated endocytosis (CME), caveolae-mediated endocytosis (CvME), and macropinocytosis [40, 42, 43]. After cellular uptake, LNPs are trafficked into early endosomes, where three main ways are possible: they can be recycled back to the membrane, degraded through fusion with lysosomes, or escape into the cytoplasm. The last route, endosomal escape, is the desired one, but it occurs typically in only 1–2% of internalized particles (mostly under 5%) [41, 44–46]. Two main mechanisms are underlying this process. One is that ionizable lipids interact with endosomal membranes to form non-bilayer structures that disrupt the membrane and release the cargo. The other is the proton sponge effect [47]. With more and more nanoparticles entering endosomes, the colloid osmotic pressure becomes higher causes osmotic swelling and finally rupture of the endosome. Improving the key factors underlying the mechanism may provide a more effective way to enhance endosomal escape efficiency. For example, cone-shaped ionizable lipids promote endosomal escape by inducing negative membrane curvature and non-bilayer phase transitions under acidic conditions, which destabilize endosomal membranes and facilitate cytosolic release of mRNA. The cone-shaped lipid shows higher endosomal escape at about 15% efficiency which is significantly higher than usual [47]. Another critical challenge in LNP development is the unintended activation of innate immunity, which can profoundly affect both the therapeutic efficacy and safety profile of nucleic acid therapeutics. LNPs and their encapsulated nucleic acid cargos are frequently recognized as foreign entities by the host’s pattern recognition receptors (PRRs) [48]. Key PRRs involved include endosomal Toll-like receptors (TLRs) [49], cytosolic retinoic acid-inducible gene I (RIG-I)-like receptors (RLRs) [50, 51], and DNA-sensing pathways such as cGAS-STING [52, 53]. For instance, mRNA delivered via LNPs can strongly activate endosomal TLR7 and TLR8 in immune cells such as plasmacytoid dendritic cells (pDCs) and monocytes [54]. This recognition triggers downstream signaling through MyD88, resulting in robust production of type I interferons (primarily IFN-α) and pro-inflammatory cytokines including TNF-α, IL-6, and IL-12. A well-known clinical example is the early mRNA vaccine platforms (e.g., some COVID-19 mRNA candidates before full optimization) [55, 56]. The high levels of TLR7/8 activation led to significant reactogenicity, manifested as fever, fatigue, myalgia, and injection-site inflammation, sometimes limiting the acceptable dose range. Beyond the nucleic acid cargo itself, the ionizable lipids used in LNPs also play a significant role in immune stimulation. Many ionizable cationic lipids, particularly those with specific headgroup chemistries, can directly interact with immune cells or disrupt cellular membranes in a way that triggers inflammatory signaling. For example, certain early-generation ionizable lipids, such as DLin-MC3-DMA or similar structures, have been reported to activate the NLRP3 inflammasome in macrophages, leading to IL-1β and IL-18 release [57]. In preclinical studies, intravenous administration of empty LNPs (without nucleic acid) still induced moderate cytokine elevation and complement activation in animal models, highlighting that the lipid components alone can contribute to immunogenicity [58–60]. Furthermore, DNA-based cargos (such as plasmid DNA or circular DNA) can engage cytosolic DNA sensors [61, 62]. The cGAS-STING pathway is particularly relevant here: upon detection of double-stranded DNA in the cytosol, cGAS produces the second messenger 2’3’-cGAMP, which activates STING [63]. This leads to TBK1-IRF3 signaling and strong type I IFN responses [64]. In the context of LNP-delivered gene editing tools (e.g., CRISPR-Cas9 mRNA + sgRNA or DNA templates), unintended activation of cGAS-STING has been linked to reduced editing efficiency and increased inflammatory side effects in some in vivo models [65, 66]. These innate immune responses can have dual consequences: while mild activation may act as an adjuvant to enhance adaptive immunity (beneficial in vaccine applications), excessive or uncontrolled activation in therapeutic settings (e.g., protein replacement therapy or gene silencing) often leads to several problems, rapid clearance of the LNPs, reduced protein expression due to translational shutdown, induction of anti-drug antibodies, and in severe cases, cytokine release syndrome or organ toxicity. In addition to the challenges previously discussed regarding LNPs, the immunogenicity of PEGylation has also become a significant obstacle, leading to various adverse reactions during LNP applications. PEGylated lipid, as noted before, can reduce clearance. It also leads to anti-PEG antibodies which may accelerate blood clearance (ABC) or even induce Complement-related pseudoallergy (CARPA). A study has found mRNA vaccines, including Moderna mRNA-1273 and Pfizer BNT162b2, can promote the induction of anti-PEG antibodies, with responses more frequently observed and more pronounced following mRNA-1273 (mean 13–17 folds after the second 100 µg dose) than BNT162b2 (1.1–1.8 folds after the second 30 µg dose). While the clinical implications remain uncertain, elevated anti-PEG antibody levels have been linked to increased systemic reactogenicity such as fever, myalgia [67–69]. Prospect of LNP: improvements in new components Strategies to overcome PEG-related Immunogenicity Some ways are founded to lower the immunogenicity reaction toward PEG. First of all, we need to figure out the factors of influencing its immunity: time intervals between first and second injection, different doses, PEG surface density and content, nanoparticles properties, administration route and encapsulated drug [70]. All of these can make an impact on the severity and time of ABC. By manipulating these parameters, the immune responses can be controlled. Another prospect is to optimize or replace PEG. One strategy centers on structural modification of the PEG moiety, for example, by substituting the methoxy terminus with alternative functional groups such as amino (–NH₂), carboxyl (–COOH), or hydroxyl (–OH) [41]. In addition, PEG itself may be replaced with other hydrophilic polymers, such as polyvinyl alcohol, polyvinylpyrrolidone, or polyacrylamide or zwitterionic polymers [71–73]. What’s more, it is reported that PEGylated liposome modified with 10% ganglioside barely induced anti-PEG IgM production and may tolerate B cells by binding inhibitory co-receptors [74, 75]. It may be explained by the conformation of PEG and the accessibility of its epitopes on the membrane. Strategies to redirect LNP tropism beyond the liver To enhance LNP accumulation at the target and reduce undesired hepatic tropism, nucleic acid delivery can be optimized through passive targeting, active targeting, and endogenous targeting [37]. Passive targeting means changing nanoparticle properties, like size, charge, and shape by the target tissue’s anatomy and physiology. For example, particles ranging from 50 to 100 nm tend to accumulate in the liver due to the fenestrated structure of hepatic sinusoids. In contrast, smaller nanoparticles (< 50 nm) are more likely to undergo renal filtration or penetrate tumor tissues through the enhanced permeability and retention (EPR) effect [39, 40, 76]. Active targeting involves modifying the surface of nanoparticles by adding ligands (such as antibodies, peptides, or nanobodies) that specifically recognize target cell receptors. This modification enhances targeting specificity and reduces off-target effects. By binding to the target receptors, it increases the cellular uptake efficiency of nanoparticles. Similar researches include using the EGFR-targeted ionizable LNP to enhance in vivo mRNA delivery to the placenta [77] and using cationic cholesterol-dependent LNP delivery to lung stem cells and heart [78]. And endogenous targeting adjusts the lipid composition or structure to exploit natural biomarker and protein coronas in the body for tissue-specific delivery. For example, Savan et al. found bile acid-containing C14-4 LNPs were able to reduce delivery to liver cells in vitro and improve delivery in T cells, B cells and so on [79]. The same effect was seen in LNPs with either C14-4 or C12-200 as the ionizable lipid, indicating that this strategy of replacing cholesterol with cholic acid could be broadly applicable [37]. Another inspiring strategy is the use of selective organ targeting (SORT) nanoparticles for systematically engineering the extracellular tissues. SORT molecules enable tissue-specific delivery of therapeutic agents by modifying LNPs with a supplemental molecule that alters their biophysical properties, particularly internal charge. Increasing the percentage of DOTAP in the LNP formulation progressively shifted the tissue targeting of the nanoparticles from the liver to the spleen and eventually to the lungs. Specifically, when DOTAP content was increased from 0% to 50%, the mRNA delivery was systematically redirected from the liver to the spleen, and then to the lung. At 50% DOTAP, lung-specific delivery was achieved, although with some reduction in activity. This conclusion is universal due to its work in 5A2-SC8 LNPs, C12-200 LNPs and DLin-MC3-DMA LNPs [80]. The research provides a flexible way to transfer nucleic acid to extra-hepatic organs. Though SORT molecules represent a novel strategy for extra-hepatic delivery, they have been reported to induce massive thrombosis and exacerbate pre-existing inflammation [81] likely due to their dependence on LNP surface charge. To overcome this limitation, Chang et al. developed a pioneering platform termed Peptide-encoded Organ-Selective Targeting (POST), which relies on a PEGylated lipid sheath. In this system, researchers synthesized a small library of short peptides composed of single amino acid motifs and conjugated them to LNPs for luciferase mRNA delivery. By monitoring luminescence in different organs, peptide sequences with specific tropism for the lung and spleen were identified. POST enables selective delivery of genetic cargos, including mRNA, antisense oligonucleotides (ASOs), and gene editing tools such as Cre mRNA and CRISPR–Cas9 to extra-hepatic organs, including the lung, spleen, placenta, and bone marrow through the alteration of peptide codes. Unlike SORT, POST functions independently of LNP charge and instead relies on corona proteins and their cognate receptors, thereby reducing the risk of charge-related side effects. Moreover, the platform incorporates an artificial intelligence–driven framework to predict peptide sequences with high organ affinity, which has been validated in vivo. Collectively, POST represents a promising strategy for the precise delivery of nucleic acid and gene editing tools. With the massive data resources and powerful analytical capabilities of artificial intelligence, organ or even tissue specific delivery may eventually become a reality [82]. Polyplexes Constitution of polyplexes Polyplexes, linking cationic polymers and negatively charged nucleic acids, are strong complements of LNP [83]. Commonly used polymers include synthetic ones such as polyethylenimine (PEI) and poly-(l-lysine)(PLL), and poly(β-amino ester) (PBAE) [84, 85]; natural polymers such as chitosan [86] and cyclodextrin derivatives [87]; and functional polymers such as poly-histidine (pHis) homopolymers and poly(ethylacrylic acid) (PEAA) (Fig. 3). PEI and PLL are usually typical options for polymers. However, their non-degradable backbone and high charge density often result in significant cytotoxicity [88]. This concern has driven the development of biodegradable synthetic polymers such as PBAE as safer alternatives. Owing to its ester backbone, PBAE exhibits degradability and thus, enhanced biocompatibility. Although PBAE exhibits instability in serum, it possesses vast structural diversity. This versatility allows proper optimization of their physicochemical and biological properties [85, 89]. Different from synthetic polymers, chitosan is natural, low toxicity but poor transduction efficiency. And functional polymers engineer materials that incorporate responsive or degradable linkages, enabling polyplexes to achieve controlled release, enhanced biocompatibility, and improved targeting. Commonly used polymers in polyplexes and cationic lipids in lipoplexes. (e) Poly(amidoamine) dendrimers (PAMAM), branched polyethyleneimine (PEI), poly(β-amino esters) (PBAE), poly(L-lysine) (PLL), and poly(lactic-co-glycolic acid) (PLGA) are among the most widely used polymeric materials for gene delivery. PAMAM and PEI possess abundant amine groups that facilitate nucleic acid condensation and promote endosomal escape via the proton sponge effect, while PBAE further introduces hydrolyzable ester linkages that enable degradation under physiological conditions; PLL and PLGA provide biocompatible and biodegradable backbones suitable for constructing stable and tunable delivery systems. (f) DOTAP, DOTMA, and DC-Cholesterol are representative cationic lipids used in lipoplex formulations, in which their permanently positively charged head groups drive electrostatic complexation with nucleic acids, while their hydrophobic tails (or the cholesterol scaffold in DC-Cholesterol) support membrane assembly and cellular uptake Structure-activity of polyplexes Structure-dependent organ targeting By altering the constitution, surface charge and size, polyplexes can exhibit distinct organ distribution. For example, cationic polymers tend to accumulate in negatively charged tissues, such as inflamed lungs. A cationic polymer called PONI-Guan (poly(oxanorborneneimide)–guanidinium polymer) can naturally accumulate in inflamed lung tissue, which becomes more negatively charged during inflammation. It can effectively deliver siRNA for gene knockdown to ease the inflammation status [90]. In addition to surface charge, particle size also significantly influences biodistribution. Smaller polyplexes generally exhibit improved tissue penetration, while larger complexes are more prone to uptake by the reticuloendothelial system (RES), particularly in the liver and spleen [91]. Studies examining polyplex size effects on biodistribution confirm this relationship. Research using siRNA polyplexes of 25 nm versus pDNA polyplexes of 162 nm showed that the smaller siRNA polyplexes penetrated significantly deeper into tumor spheroids after 24 h, while larger pDNA polyplexes remained primarily in the rim [92]. Surface modification such as PEGylation can further influence these interactions. Furthermore, the chemical structure of the polymer can dictate organ tropism. For instance, fluorinated amphiphilic PBAE nanoparticles exhibit strong lung tropism following systemic administration [93]. Similarly, poly(amine-co-ester) (PACE) nanoparticles with different compositions showed that polymer chemistry directly influences tissue tropism for nucleic acid delivery. By varying polymer composition (e.g., PEGylation and surface functionalization) while controlling nanoparticle size, the authors observed distinct biodistribution patterns across organs and cell types, indicating that nanoparticle fate cannot be predicted by size alone but is strongly influenced by polymer structure [94]. Structure dependent toxicity and efficiency Except for different constitutions (such as subunits), structural variations also play a decisive role in the biological activities of polyplexes. Even subtle changes in topology, combining sites, or particle size can significantly influence stability, cytotoxicity, and transfection efficiency [83]. Linear polymers generally exhibit greater chain flexibility, whereas branched polymers provide a higher density of functional groups for nucleic acid binding [95, 96]. However, branched polymers such as branched PEI are often associated with higher cytotoxicity, likely due to their higher density of primary amines and stronger interactions with cellular membranes, which are associated with apoptosis induction [97, 98]. Beyond conventional linear and branched architectures, novel supramolecular configurations have been explored. For instance, Cheng et al. reported polycations with a sunflower-like configuration based on 2-dimethylaminoethyl methacrylate, which exhibited a higher IC50 value and more importantly, higher gene transduction efficiency compared to linear and comb-like analogues [99]. At first, the synthetic polymers, PEI, are hard to widely use due to their high cationic density, non-degradable backbone and molecular weight which result in toxicity. That’s why modifications are necessary for its wider use. A novel degradable PEI with acid-labile imine linkers shows equal transfection efficiency but shorter halflife at pH7.4 which means less toxic than traditional PEI [100]. Moreover, using stimuli-responsive elements and shieldings such as redox-sensitive linker, pH-sensitive linker, polysaccharide shielding and hydrophobic shielding may be beneficial for improving biodegradability, thus mitigating its toxicity. Promising approaches such as polypropylenimine diarylate ester, ketal, are verified to improve tolerability in B16-F10 [101] and NIH3-T3 [102]. PEG is also introduced to PEI to reduce serum adsorption and cytotoxicity. The PEGylated-PEI shows better condensation and complexation properties. And the transfection efficiency was increased 10 fold than the usual PEI with a longer circulation time [103]. Another method to reduce its toxicity is to shield PEI with polysaccharide and hydrophobic moieties. Though most attempts seem to have tiny effect. Arg-His and Arg-Ser sequences coupled with PEI have achieved higher transfection activity [104]. Another attempt is about hydrophobic shielding. PEGylated-PEI shows improved safety but unsatisfactory transfection efficiency for its cellular uptake and hindered endosomal escape. Hydroxyethyl starch with PEI reduces aggregation and phagocytosis as well as enhances the uptake and intracellular release [105]. The structure modifications of PEI recently aimed to mitigate its cytotoxicity which comes at unexpectedly lower efficiency, leaving a narrow therapeutic window for efficient and safe gene delivery. Possible solutions may be formed from precise regulation of hybrid strategies combining different topology polymers with smart responsive elements. In this case, a more balanced, flexible and widely applicable vector is within reach. Applications of polyplexes Cancer therapy Polyplexes are mainly applied in cancer therapy, genetic disease treatment and so on. Nowadays, with the increasingly wide use of LNP, polyplexes seem to lose its light in the nucleic acid transduction market. In fact, recently, researches on polyplexes have not come to a halt but has rather undergone continuous optimization and diversification, focusing on enhancing biocompatibility, targeting specificity, and controlled release properties. In oncology, polyplexes mainly function through enhancing immune reactions and regulating transcription of cancer-related factors in signaling pathways. Structurally dynamic polyplexes, composed mainly of two poly(ethylene glycol)-(glycine–lysine)\10[PEG-(GK)10] chains complexed with antisense oligonucleotides (ASOs), were designed to deplete transforming growth factor-β1 (TGF-β1) in sentinel lymph nodes and thereby rejuvenate cytotoxic CD8⁺ T cells. In murine breast cancer models, this approach significantly reduced postoperative tumor recurrence and lung metastasis. Importantly, by optimizing the PEG chain length, the polyplexes achieved selective sentinel lymph nodes accumulation while minimizing off-target distribution, which improved both therapeutic efficacy and safety [106–108]. Interleukin-12 (IL-12) is a potent pro-inflammatory cytokine that stimulates cytotoxic T lymphocyte activity and enhances antitumor immunity. To exploit this property, researchers developed esterase-responsive polyplexes by complexing IL-12 plasmids (pIL12) with PQDEA, enabling efficient transfection of both tumor cells and tumor-associated macrophages (TAMs). To further improve tumor targeting, these polyplexes were coated with DSPE-PEG-AEAA, yielding APEG-LPs/pIL12, which selectively accumulate in TAMs and tumor cells which boost the anti-tumor efficiency and reduce undesired accumulation. In three TAM-rich tumor models, KPC pancreatic carcinoma, BPD6 melanoma, and 4T1 breast cancer, APEG-LPs/pIL12 significantly suppressed tumor progression, doubled median survival compared to PBS, and induced durable antitumor responses. Treatment markedly increased intratumoral IL-12 expression at both mRNA and protein levels, remodeled the TME by repolarizing TAMs toward the M1 phenotype, enhancing CD8⁺ T-cell infiltration, and reducing immunosuppressive populations. Importantly, this system achieved robust efficacy with minimal systemic toxicity, highlighting its promise as a nonviral platform for cancer immunotherapy. What’s more, other immune or signaling related nucleic acids, gene editing tools or drugs are also hopeful to be delivered by the system [109]. PONI-Guan75-Zwit25 [72], a zwitterionic copolymer siRNA polyplex, has been shown to efficiently deliver siRNA targeting STAT3. Due to hydration shell formed by stealth Zwit and the dense core formed by Guan and siRNA spontaneously, PONI-Guan75-Zwit25 evaded clearance by monocytes. STAT3 is expressed both in macrophages and cancer cells and alters the relationship between TME and TAM which indicates its potential as a target. This system achieved more than 80% knockdown efficiency at relatively low siRNA doses (0.14–0.28 mg/kg), markedly reducing off-target effects. The capacities of migration and growth were significantly inhibited by over 80% while no hemolysis was observed. The results showed great efficiency with low systemic toxicity, though further in vivo validation is still required. Similarly, an injectable thermosensitive hydrogel system has been developed to locally deliver siRNA-STAT3 polyplexes for advanced ovarian cancer therapy recently. In this system, siRNA was first complexed with a cationic diblock copolymer to form ~ 25 nm polyplexes and subsequently embedded into a pNIPAM–PEG-based thermosensitive hydrogel. The system enables in situ gelation and sustained release following intraperitoneal administration [110], illustrating a promising strategy for integrating polyplexes with biomaterial-based delivery platforms. Deng et al. [111] developed a prodrug polyplex system termed PAC, which integrates CRISPR/dCas9-KRAB to downregulate PD-L1 with the epigenetic inhibitor azacytidine (AZA) to achieve synergistic effects. AZA upregulates MHC-I expression, induces tumor-associated antigen (TAA) presentation, and promotes dendritic cells (DC) maturation, while CRISPR/dCas9-KRAB effectively suppresses PD-L1 expression to achieve immune checkpoint blockade. Together, these mechanisms transform an immunologically “cold tumor” into a “hot tumor,” thereby enhancing immune recognition. In vitro, PAC was rapidly internalized by 4T1 and B16F10 tumor cells and significantly promoted immune activation through DC maturation. In vivo, PAC accumulated in tumor-draining lymph nodes (TDLNs) and achieved the smallest tumor volume with an inhibitory rate of 83.51%, exceeding PBS, liposome, AZA, or CRISPR/dCas9 monotherapy. Further analysis revealed that PAC treatment enlarged TDLNs with higher DC infiltration, while within the TME, CD8⁺ and CD4⁺ T cells were elevated and Tregs were reduced, indicating robust immune remodeling. Moreover, the PAC system addresses the issue of the poor in vivo biodistribution of AZA, which has long limited its efficacy against solid tumors. This study highlights a synergistic polyplex-based platform that integrates gene therapy, immunotherapy, and epigenetic modulation, presenting robust efficacy both in vitro and in vivo. Although the current investigation focuses on triple-negative breast cancer and melanoma, both of which relate to strong immunological responses, the strategy may also be applicable to a wider range of immune associated solid tumors. Polyplexes have been employed to deliver a wide range of therapeutic cargos for cancer gene therapy, including plasmid DNA (pDNA) [112], ASO [106], siRNA [72, 113, 114], miRNA [115], mRNA [116, 117] and CRISPR-Cas9 [111]. These systems have been designed to modulate diverse oncogenic pathways [114], silence immune checkpoints such as PD-L1 through KRAB-mediated repression [111], target tumor microenvironmental molecules like CXCR4 [112] and regulate epigenetic factors such as DNMTs [111]. With versatile possible targets and stable transduction efficiency, polyplexes show unique strength in programmability, biocompatibility, and delivery precision which is difficult to achieve with traditional vectors. Genetic disease Duchenne muscular dystrophy (DMD) Duchenne muscular dystrophy [118] (DMD) is an X-linked inherited disorder caused by deletions or mutations in the dystrophin gene, resulting in fragile muscle fibers and progressive muscle degeneration. Polyplexes have been investigated to overcome this defect through different approaches, including the delivery of ASO for exon skipping [119, 120] and CRISPR-Cas9 [121] for gene editing. Exon-skipping oligonucleotides are promising therapeutic candidates for DMD; however, their in vivo application is hindered by limited transduction efficiency in muscle tissue [119]. To improve efficacy, a polyplex formulation composed of PEI–PEG copolymers combined with the cell-penetrating peptide (TAT) and colloidal gold (CG) demonstrated markedly enhanced outcomes, achieving a six-fold increase in dystrophin restoration compared to ASOs alone and up to 30% dystrophin expression as confirmed by Western blot [122]. Polyplex nanomicelles formed by proliferator-activated receptor gamma coactivator-1 alpha (PGC-1α) mRNA and a synthetic block copolymer of PEG and polyamino acid (poly{N-[N’-(2-aminoethyl)−2-aminoethyl]aspartamide}, PEG-PAsp(DET)) yielded sustained protein expression with reduced immune activation and tissue damage [123]. Given that PGC-1α is a key regulator of skeletal muscle resistance to damage and is downregulated in dystrophic mdx mice, its overexpression has been associated with functional improvements [124]. Notably, after repeated eccentric contraction (ECC), a standard protocol to evaluate muscle fragility, PGC-1α polyplex nanomicelles treatment significantly preserved isometric torque, with reductions in functional decline exceeding 30% compared with saline, nanomicelles alone, and even LNP-PGC-1α group (which achieved ~ 20% reduction). Moreover, ECC-induced myofiber injury, as assessed by Evans blue dye uptake, was lowest in the nanomicelle-PGC-1α group (10.26 ± 0.93%). PGC-1α protein remained detectable at both day 2 and day 7 after injection, whereas LNP-delivered PGC-1α group declined to near baseline by day 7. Consistently, polyplex delivery also enhanced expression of PGC-1α-related genes such as Igf1 (muscle hypertrophy), Glut4 (metabolism), and VEGF-A (angiogenesis), while suppressing inflammatory markers including IL-6 (p = 0.0171 vs. saline), which were absent in the LNP-PGC-1α group. Polyplexes have more controlled endosomal escape by proton effect and avoid hepatic tropism which may account for its better performance than LNP in DMD. One of the major challenges in CRISPR-based therapy for DMD lies in the efficient co-delivery of Cas9 RNP and donor DNA templates required for homology-directed repair. To address this issue, recently, Xie et al. developed a pH-responsive amphiphilic polymer nanoparticle (mPEG-PC7A) capable of accommodating the heterogeneous physicochemical properties of protein–RNA complexes and nucleic acids through combined hydrophobic and electrostatic interactions. In a DMD (mdx) mouse model, intramuscular administration of HDR-loaded nanoparticles restored dystrophin expression across muscle tissues and significantly improved muscle function, demonstrating a ~ 3-fold increase in hanging time compared to controls [125]. Cystic fibrosis (CF) Cystic Fibrosis (CF), caused by mutations (mostly ΔF508) in the cystic fibrosis transmembrane conductance regulator gene, results in viscous mucus and chronic, recurrent bacterial infection in lungs, gastrointestinal tract and reproductive system [126, 127]. Direct inhalation delivery of CFTR pDNA or mRNA is a promising approach for CF, however, the viscous mucus impedes particle penetration [128]. Another problem for gene delivery in CF is that the CpG islands in plasmid trigger immune responses which may result in undesired immune stimulation leading to reduced transfection efficiency [129]. Polymer composition and topology can be flexibly tuned, which enables polyplex as a potential candidate for CF. In order to enhance the penetration and efficiency, lyophilized and powderized PEI/pDNA polyplexes are formulated as powder aerosols. By inhalation administration, Pfeifer et al. demonstrated robust gene expression and physiological response in mouse models which verified dry powder aerosol polyplexes as a safe and durable non-viral delivery strategy [130]. The branched type of Poly (β-amino esters) (PAEs) with E7 endcap (HPAE3) showed higher transfection efficiency in CFBE14o-cells than other structures with endcap groups with statistical significance (P < 0.0001) [131]. Combined with a CpG-depleted CFTR plasmid to reduce immune responses and human promoter human elongation factor 1α (hEF1α) to prolong CFTR expression, the HPAE3/CpG-depleted CFTR complex was formed. After 48 h transfection, a 25 fold increase in CFTR levels compared with healthy human bronchial epithelial cells was observed. Moreover, the in vitro experiments showed 35.28% recovery of CFTR function in HPAE3/CpG-depleted CFTR plasmid polyplex system compared to 30.08% recovery in the healthy group which demonstrated its capacity in restoring CFTR protein function. HPAE3 polyplexes not only exhibit high transfection efficiency and low cytotoxicity but also enable reproducible large-scale manufacturing and compatibility with industrial formulation processes such as lyophilization and spray drying which is critical for inhalation administration for CF therapy. However, safety and actual CFTR expression in lungs of HPAE3/CpG-depleted CFTR plasmid polyplex system remain unknown due to the absence of in vivo validation in animal models. However, some obstacles are challenging for the CF use of polyplexes. The dense and dehydrated mucus acts as a formidable physical barrier to gene delivery. The viscous mucus matrix, together with the limited pulmonary deposition efficiency of inhaled particles, restricts the access of polyplexes to airway epithelial cells and substantially reduces transfection outcomes [131]. Beyond the mucus barrier, the chronically inflamed airway environment of CF patients poses additional hurdles for nucleic acid delivery [132]. Persistent inflammation and immune cell infiltration enlarge the recognition and clearance of exogenous particles, diminishing the therapeutic window. The nucleic acid delivery therapy may aggravate inflammation and result in low transduction efficiency. In this case, CpG-depleted plasmids [129, 131] and low charge density polymers capable of minimizing immune activation have become crucial design elements for polyplexes in CF therapy. The treatment of CF requires long-term and repeated inhalation administration [128, 132]. Accordingly, polyplexes must endure nebulization or spray-drying processes without structural disintegration or loss of bioactivity and maintain uniform particle size distribution to ensure efficient lung deposition. Although scalable HPAE systems [131] and sugar-based lyoprotectants such as lactose [130] have emerged as promising approaches to improve formulation stability, comprehensive data on long-term inhalation safety and toxicology are still lacking. Despite promising in vitro performance, most CF gene therapy studies using polyplexes have not progressed beyond the cell culture stage [130, 131]. As a result, authentic airway dynamics, immune responses, and the durability of transgene expression in vivo remain difficult to predict, indicating the need for animal models and studies to bridge the gap from bench to the bedside. Other applications of polyplexes Beyond the diseases discussed above, polyplexes have also been investigated across a broad spectrum of other pathological conditions, ranging from genetic and neurodegenerative disorders [133] to inflammatory [90, 134, 135] and ocular diseases [136]. In neurodegenerative ataxias and Parkinson’s disease, polyplexes have been employed to deliver neuroprotective genes (neurotrophin factor gene) [137] or RNA molecules (α-SNCA siRNA) [138] that modulate neuronal survival pathways. Similarly, in retinitis pigmentosa, polyplexes facilitate the transfer of therapeutic genes to retinal cells, aiming to restore or preserve photoreceptor function [136, 139]. In hematologic disorders such as hemophilia, polyplexes have been explored for the delivery of functional coagulation factor genes, providing a safer and potentially repeatable strategy compared with viral vectors [140]. Moreover, in autoimmune diseases like rheumatoid arthritis, polyplexes capable of delivering anti-inflammatory siRNA (anti-TNF-α siRNA) have demonstrated efficacy in rheumatoid arthritis lesions within affected tissues [135, 141]. Recent advances highlight the remarkable tunability of polyplexes. By incorporating different structural motifs, these systems can be endowed with diverse functionalities, such as ROS-responsiveness [111, 142] or enhanced targeting specificity [138]. Their cargo capacity is likewise versatile, ranging from pDNA and antisense oligonucleotides (ASOs) to siRNA and miRNA. Most importantly, the applications of polyplexes in oncology span targeted therapy, immunotherapy, and chemotherapy. The unique appeal of polyplexes lies in their modular design, which enables the integration of distinct therapeutic modalities and the co-targeting of synergistic pathways to achieve enhanced antitumor efficacy. In addition, the route of administration can markedly influence therapeutic outcomes: For pulmonary diseases, inhalation or nebulization provides direct access to the airway epithelium while avoiding systemic dilution. Accordingly, aerosolized pDNA/PEI polyplexes formulated with CpG-free plasmids have demonstrated sustained transgene expression in the lung upon repeated administration, as the removal of immunostimulatory CpG motifs mitigates TLR9-mediated inflammation [129, 143]. Hyperbranched PBAE polyplexes enabled efficient inhaled mRNA delivery to lung epithelium for the treatment of Silicosis [144]. Though inhalation is suitable for lung diseases, mucus barriers on the lung epithelium remain a major challenge [145]. Intratumoral injection is particularly attractive for accessible or locally advanced solid tumors, because it can achieve a high local concentration of nucleic acids while reducing systemic exposure and off-target toxicity. This route is especially valuable for polyplexes, whose systemic administration is often limited by serum instability, nonspecific interactions, and insufficient tumor accumulation. Local intratumoral administration of targeted PEI-PEG polyplexes significantly enhanced melanoma transfection and improved the antitumor efficacy of HSVtk/ganciclovir therapy relative to non-targeted controls. In another example, a thermosensitive injectable polyplex hydrogel carrying cyclin B1 siRNA produced sustained local release after a single intratumoral injection, leading to antitumor activity for up to 4 weeks. Notably, oral administration has long been considered challenging due to the harsh gastrointestinal environment and multiple biological barriers. However, phenylboronic acid (PBA)-functionalized polyplexes have been engineered to maintain structural stability under acidic gastric conditions, facilitate transport across the mucus layer via ligand-like interactions, and enable efficient intracellular release through ATP-responsive mechanisms. Such breakthroughs broaden the applications of polyplexes and make polyplexes more clinically friendly. By flexibly integrating therapeutic targets, receptors, and delivery strategies, polyplexes hold great promise for addressing some of the most troubling challenges in oncology, such as optimizing the TME and enabling personalized precision medicine guided by next-generation sequencing (NGS) [146]. All in all, polyplexes offer significant advantages in nucleic acid delivery due to their structural tunability and robust electrostatic interactions [147]. Specifically, polymers such as PEI and histidine-rich derivatives enhance intracellular delivery by triggering the proton sponge effect, thereby promoting endosomal escape. However, several intrinsic hurdles impede their clinical translation such as cytotoxicity and colloidal stability in physiological fluids. The nonspecific interactions with serum proteins such as albumin typically lead to rapid systemic clearance and off-target sequestration [148]. The relative impact of these attributes is highly context-dependent. In oncology, where transient transgene expression via localized administration may suffice, the high transfection efficiency of polyplexes is a decisive benefit [149, 150]. Conversely, for inherited genetic disorders necessitating systemic delivery and sustained expression, stringent requirements for biocompatibility, serum stability, and manufacturing reproducibility remain the critical bottlenecks [151]. Lipoplexes Lipoplexes, generated by the spontaneous electrostatic association of cationic lipids with nucleic acids, represent one of the earliest and most widely explored formulations for non-viral gene delivery. A typical lipoplex formulation consists of permanently cationic lipids (e.g., DOTAP or DOTMA), helper lipids such as DOPE or cholesterol, and the nucleic acid cargo. Nevertheless, compared with LNP, their in vivo stability and pharmacokinetic control remain limited. Relative to polyplexes, their tunability and stability require further optimization. Lipoplexes are more like a bold try for lipid-based nucleic acid delivery, and paved the way for the rationally engineered LNP systems that dominate today’s clinical landscape. Owing to their structural simplicity and ability to facilitate cellular uptake through membrane fusion and endosomal escape, lipoplexes have been applied across cancer vaccines, cancer gene therapy, and correction of genetic diseases. Among these applications, the mRNA-lipoplex vaccines for pancreatic cancer are representative. Personalized RNA neoantigen vaccines (BNT122/RO7198457) based on lipoplexes (autogene cevumeran) for pancreatic ductal adenocarcinoma (PDAC) are able to stimulate de novo, high-magnitude polyclonal CD8⁺ responses under the fixed sequence: atezolizumab (anti-PD1), autogene cevumeran, mFOLFIRINOX [152]. Compared to patients without vaccine-expanded T cells (13.4 months), patients with vaccine expanded T cells (not reached) prolonged median recurrence-free survival (P = 0.003). In the following trial, the lipoplex vaccine validated durability and biological relevance, with vaccine induced clones persisting for years (average lifespans 7.7 years) [153]. Next, the lipoplex vaccine moves to randomized validation (IMCODE-003, NCT05968326), while efforts focus on raising responder rates, limiting escape (broader epitope coverage and spreading), and optimizing sequencing with checkpoint blockade and chemotherapy. With the trials being carried out steadily, lipoplex-based mRNA vaccines are hopeful to transform from bench to bedside, leading to wider application. Beyond clinical applications, lipoplexes have long served as foundational tools in laboratory. Representative benchmark transfection reagents such as Lipofectamine 2000 form lipoplexes with nucleic acids. Lipofectamine 2000 is composed primarily of a permanently charged cationic lipid (DOSPA, 2,3-dioleyloxy-N-[2-(sperminecarboxamido)ethyl]-N, N-dimethyl-1-propaniminium trifluoroacetate) and a neutral helper lipid (DOPE), typically formulated at an approximate 3:1 ratio [154, 155]. Lipofectamine 2000 enables effective transfection through sequential steps which make it an important tool in laboratory [156]. Compared with viral vectors, Lipofectamine 2000 allows the delivery of larger nucleic acid cargos in vitro. Compared to physical delivery methods such as electroporation, Lipofectamine 2000 does not require specialized instrumentation. All of these make Lipofectamine 2000 a widely used transfection reagent for in vitro gene delivery studies. Exosome Exosomes are endosomal origin nanoscale extracellular vesicles (30–100 nm in diameter) that are secreted by nearly all cell types (hematopoietic cells, tumor cells and so on) [157]. They are produced by inward budding of multivesicular bodies (MVBs) and then released into the extracellular after fusion of MVBs with the plasma membrane [158]. Exosomes contain complex cargoes of proteins, lipids, and nucleic acids that are fundamental in intercellular communication, immune regulation, and disease development, including cancer and neurodegenerative diseases [159]. However, the definition of exosomes remains controversial because of the lack of unique molecular markers and the overlapping size range with other extracellular vesicles like microvesicles (100–1000 nm in diameter) which makes exosomes more like a functional term. And in this part, we will only discuss its application as a vector in medical use considering associated articles. Applications of exosomes Exosomes play crucial roles in biological, pathogenic and medical use. As a natural vector, exosomes show particular promise in gene therapy, due to their nice biological compatibility [157], high stability [160] and natural targeting [161]. Besides, exosomes can be isolated from a patient’s own cells, such as stem cells or immune cells, which minimizes the risk of immune rejection and makes them suited for repeated administration and personalized precision medicine [162]. Valadi et al. first revealed that exosomes are not simple carriers of cellular debris but rather active carriers of genetic information. They found that exosomes carry unique sets of mRNA and microRNA that are selectively packed and can be functionally transferred to recipient cells, where the mRNAs can be translated and the microRNAs can modulate gene expression [163]. This finding radically altered the view of exosomes and positioned them as significant mediators of intercellular communication with potential functions in gene regulation, immunity, and disease development [164]. Based on this, researchers soon began to explore whether exosomes could be harnessed not only as endogenous messengers but also as exogenous nucleic acids. Wahlgren et al. Found that human plasma-derived exosomes can act as natural nanocarriers for exogenous siRNA delivery. Researchers innovatively employed exosomes derived from HeLa and lung carcinoma cells, which exhibited typical exosomal features including 50 nm vesicular morphology and enrichment of CD9, CD63, CD81, Hsc70, and TSG101 which are characteristic of exosomes. This work established a crucial proof-of-concept for using self-derived vesicles as gene delivery vectors, addressing a major limitation in RNAi therapy: the lack of biocompatible and efficient carriers. By successfully loading siRNA into exosomes through electroporation and confirming their ability to deliver functional RNA to human monocytes and lymphocytes, the authors provided the first evidence that circulating immune cells can uptake exogenous siRNA via exosomes. This finding not only verified the biocompatibility and functional capability of plasma exosomes but also suggested a physiological route for systemic RNA transport and immune modulation. Beyond its technical innovation, the study laid the conceptual foundation for later therapeutic exosomes platforms, demonstrating that endogenous vesicles could circumvent the toxicity and immunogenicity associated with synthetic or viral vectors, thereby creating new ways for non-viral gene therapy [165]. Cancer therapy Exosomes are ideal vectors for the flexibility to target different cells and load various cargoes. By engineering related protein in the donor cells, the target peptides are placed on the membrane of exosomes which can navigate to transfect cancer cells with specific molecules. In an EGFR-expressing breast cancer model, researchers engineered donor cells (HEK293) to express GE11 peptide which can combine with EGFR molecules specifically. Let-7a miRNA introduced in exosomes by electroporation showed the best efficacy. The let-7a containing GE11 reached tumor tissues in vivo three times that of the control group which shows great ability for targeting. After injecting let-7a containing GE11 positive exosomes in RAG2–/– mice in vein, it was found obvious suppression in tumor growth (n = 6, p < 0.01) [166]. The outstanding targeting ability and efficiency to inhibit tumor growth show exosomes’ potential as a promising vector with low immunogenicity. Besides, due to the natural membrane protein CD47-mediated protection, exosomes reduced the phagocytosis by monocytes and macrophages which can protect cargoes from degradation and further improve the transduction efficiency. Different from synthetic materials like polyplexes and LNP, exosomes have a special talent in avoiding clearance. Similarly, KRAS, an oncogenic gene, is main driven factor and potential target in PDAC though the efficiency and safety in transporting the drugs to cancer remains unsolved. In order to knockdown KRAS in cells, researchers selected normal fibroblast-like mesenchymal cells exosomes to introduce KRAS siRNA and short hairpin RNA (shRNA) by electroporation in it. Compared with liposomes, exosomes showed less clearance and longer circulation time. Cellular uptake visualized by luciferase and gene silencing only occurred in KRAS-positive cancer cells which verified its targeting and silencing efficiency [167]. Except for gene therapy, exosomes are also candidates for cancer vaccines. As natural vesicles, exosomes have little immunogenicity and high biocompatibility compared with LNP and polyplexes. In the application of cancer vaccine, dendritic cell-derived exosome vaccines (DEX) and tumor-derived exosomes (TEX) vaccines are the 2 main kinds. In recent years, DEX has been well discussed. With MHC-I, MHC-II and some co-stimulatory molecules like CD80,CD40 and CD86 on the surface, exosomes stimulate the expansion of tumor-specific cytotoxic T Lymphocytes (CTL), reactivate anergic T cells in TME and co-opt NK-mediated cytotoxicity [168, 169]. The investigation of DEX started in 1998. Tumor peptide loaded DC and DEX were examined as 2 groups in mouse model. Contrary to the traditional views, the DC group merely affects the growth of the tumor as DEX group inhibited the growth of tumor and induced complete tumor regression in 60% mice [170]. This result may be related to more MHC molecules on DEX membrane compared to DCs. After that, more clinical trials were executed. A phase I study of DEX loaded with MAGE antigens in 13 patients with advanced NSCLC reported in 2005. With mild adverse effects, the survival of patients after the first DEX was 52–665 days. 1/3 patients detected specific T cells responses while 2/4 patients detected NK lytic activity [171]. The finding indicated the toleration of DEX in NSCLC patients and its underlying mechanism for triggering T cells activation and NK lytic activity. In 2015, a phase 2 clinic trial were conducted using IFN-γ-DEX as maintenance therapy for advanced NSCLC patients after first-line chemotherapy. IFN-γcan induce the maturation of DC cells and enable DEX with higher immunostimulatory by upregulating MHCII, CD40, CD80, BAG6 and HSP70 [172–174]. Among 22 patients, the IFN-γ-DEX significantly ameliorated the PFS and OS with 7 patients experiencing stabilization over 4 months. T cells responses were not triggered by IFN-γ-DEX while the function of NK cells was boosted by the combination of Natural killer protein 30 (NKp30) and BAG6 antigens [175]. Further analyses revealed a positive association between the MHC class II content of Dex and the magnitude of NKp30-dependent NK cell activation. In parallel, the expression level of BAG6, a known ligand for NKp30, was found to correlate with MHC-II abundance in the Dex. The finding indicates that the content of BAG6 in serum may be decisive to the degree of immune responses triggered by DEX and influence in the PFS and OS of the patients. Except for triggering immune responses, exosomes also stand out for the stable composition and phenotype which means they have little chance to be influenced by TME or other immune cells. Sometimes, DC will switch to immunosuppressive subtype as the tumor progression [176] while dex will be little influenced. Moreover, exosomes are reported to influence TME by promoting cytokines and inhibiting immunosuppressvie functions of stromal cells. DEX not only stimulates T cell longitudinally, but also cross-dressing DC cells by transferring antigen transversely and further amplifies the immune response. Moreover, a recent study indicated DEX may stimulate T cells through tumor cells. The incorporation of DEX and tumor cells will turn tumor cells to an attractive target for immune cells [177]. As a candidate for cancer vaccine, DEX are also convenient to store. Storage for − 80℃ at 6 months can still preserve viability [178, 179]. Though the pre-clinic articles and reviews of DEX are abundant, the clinical applications still face huge resistance due to unsatisfactory results. Some reliable inferences have been made to explain the phenomenon. Importantly, the insufficient stimulation of immune systems accounts for the results, especially the adaptive immune. Most of the DEX trials can not stimulate T cells fully, only with a MHC-related stimulation of NK-cells [175]. The explanations of insufficient stimulation were diverse. At a molecular level, the MHC-I/II restricted peptides are unable to trigger T cells. That’s why detecting mutated neoantigens is of great importance. By loading neo-antigens, DEX possibly trigger stronger stimulation to T cells. Decorating the exosome with potent immune modulators may enhance the immunogenicity as well. CpG-ODN-binding peptides (CpG-oligodeoxynucleotides) can be attached to the exosome surface to co-deliver the exosome’s native tumor antigens alongside the potent TLR9 agonist, enhancing antigen presentation by DC. On patient level, taking the clinic trials into consideration, most of the trials incorporated patients with advanced malignancies and already receiving first-line therapy which influences the outcomes to a great extent. The immunosuppressive TME (Treg, MDSC, etc.) in advanced patients remains a major obstacle in triggering immune responses. The number and heterogeneity of patients aren’t sufficient for reliable verification. That’s why more clinic trials with pre-selection standards are urgently needed for further exploration of DEX. And most importantly, the accurate molecule composition of DEX is still unknown. As a multi-functional molecule ranging from 30 to 150 nm, the exact constitution of DEX still has some questions. Is there PD-L1 or PD-L2 on the DEX membrane? Do any other molecules we don’t know affect the immune system? Questions need to be figured out in the future studies. Neurological disorders Gene therapy is also a important part in neurological disorders. However, with the block of the BBB, the penetration of drugs to the Central Nervous System (CNS) remains a huge problem. By engineering self-derived DC to express Lamp2b, Lamp2b on the membrane of DC-derived exosomes can specifically combine with the nicotinic acetylcholine receptor (nAchR) of the neurons. Thus, the exosomes can penetrate BBB and deliver BACE (a possible target for Alzheimer’s disease) -siRNA to neurons. With 60% knockdown of mRNA and 62% reduction of protein, exosomes showed their ability to penetrate BBB and efficiency in silencing target genes [180, 181]. With non-specific uptake in other tissues and non T cell proliferation observed, the safety of exosome-delivered BACE siRNA can be proven. The existence of exosomes makes the delivery to CNS possible. And with more and more targets appearing, the diseases hard to cure may have new options for their therapies. Kojima et al. established a synthetic biology-based platform termed EXOtic (EXOsomal Transfer Into Cells), enabling programmable production of “designer exosomes” with enhanced biogenesis, selective mRNA loading, cytosolic delivery, and brain targeting. Owing to their ability to cross the BBB with the help of RVG-Lamp2b on the membrane, exosomes enable the precise delivery of therapeutic agents to previously inaccessible lesions within the CNS. By implanting engineered cells that continuously secreted catalase mRNA-loaded exosomes, the study achieved significant decrease of neuroinflammation and neurotoxicity in Parkinson’s disease models [182]. Other diseases For cardiovascular diseases, mesenchymal stem cell (MSC)-derived exosomes promote cardiac regeneration and angiogenesis after myocardial infarction by transferring functional miR-21 and other cardioprotective molecules to recipient cardiomyocytes and endothelial cells [183]. For regenerative medicine and tissue repair, Adipose-derived stem cell (ADSC) exosomes accelerate diabetic wound healing by delivering IL-33 from macrophages to boost angiogenesis and reduce inflammation [184]. For autoimmune diseases, MSC-derived exosomes exert potent immunomodulatory effects by suppressing Th17 differentiation and pro-inflammatory cytokines like IL-17, TNF-α, thereby alleviating inflammation in collagen-induced arthritis models [185]. The gene therapy therapeutic field of exosomes spans a wide range of diseases: from cancer and regeneration medicine to neurological and cardiovascular systems, indicating their potential as natural cell-free vectors. Challenges of exosomes Exosome therapies continue to be held back by essential challenges. The basic challenge is the insufficient knowledge we have about the exosomes. The field still faces fundamental unknowns concerning their heterogeneity, selective cargo loading, cellular uptake, cytosolic release, and in vivo fate. These unresolved questions continue to limit the translation of exosomes into standardized therapeutic platforms. Without a comprehensive understanding of exosomes, purifying exosomes is also challenging. Another challenge lies in the efficient loading of therapeutic cargoes. Compared with polyplexes or LNPs, the encapsulation efficiency of nucleic acids or small-molecule drugs into exosomes remains low and highly variable, largely depending on the loading strategy [186, 187]. Passive loading during exosome biogenesis often results in insufficient cargo incorporation, while post-isolation methods such as electroporation or sonication may compromise vesicle integrity or lead to cargo aggregation. The difficulties and uncertainty in loading cargoes pose challenges for applications that require high-dose delivery. In the delivery aspect, the efficiency of exosome-mediated cargo transfer remains limited by multiple biological barriers. The negatively charged phospholipid bilayers of both the exosomal membrane and the cellular plasma membrane create strong electrostatic repulsion, which restricts direct interaction and internalization. As a result, most exosomes are taken up through endocytic pathways rather than direct membrane fusion, leading to inefficient cellular uptake and partial sequestration within endosomes. Moreover, once internalized, only a small fraction of exosomes can successfully escape from the endo/lysosomal system into the cytosol. This low cytosolic release efficiency severely restricts the functional delivery of therapeutic payloads such as siRNA, mRNA, and proteins, since the majority of exosomal cargoes are degraded in acidic compartments before reaching their cytoplasmic or nuclear targets. Consequently, overcoming these barriers, enhancing cellular uptake and promoting endosomal escape, has become a key focus in engineering exosomes for efficient gene and drug delivery. Moreover, the absence of standardized large-scale manufacturing remains a major challenge for exosome production. Although several industrial-scale GMP facilities for exosomes have been established, the field is still far from achieving fully standardized, reproducible, and available production for clinical use. Prospect of exosomes as vectors In recent years, protein engineering of exosomes has emerged as a central strategy to overcome key limitations in cargo loading, pharmacokinetics, and targeting specificity. By genetically modifying exosomal membrane proteins or directly engineering the vesicle surface, exosomes can be transformed into programmable delivery platforms. A major bottleneck in the intracellular delivery of exosome cargoes is the vesicle’s tendency to be trafficked to the late endosome or lysosome uptake. To solve this, the incorporation of Fusogenic peptides, such as GALA or INF7, is a critical strategy. These peptides are designed to undergo a conformational change in the acidic environment of the endosome, promoting the destabilization and fusion of the exosomal and endosomal membranes. This “membrane bursting” effect facilitates endosomal escape, efficiently boosting the therapeutic cargo into the recipient cell’s cytoplasm. As the purification of exosomes, using the culture medium depleted of bovine serum may help avoid the exogenous exosomes at first. Common-used techniques to purify exosomes include density gradient ultracentrifugation, magnetic activated cell sorting (MACS), size-exclusive chromatography (SEC) and chemical precipitation (Exoquick from System Bioscience Inc) [188–190]. Despite the rapid development of isolation techniques, the purification of exosomes remains intrinsically challenging due to their nanoscale heterogeneity, complex origins and the vague definition and borderline between exosomes and EV. Achieving programmable tropism is the main goal for the clinical translation of exosome therapeutics. Current methodologies can be broadly categorized into two major approaches: genetic manipulation of the producer cells for endogenous display, and direct modification of the exosome membrane. The cell manipulation, leveraging the cell’s natural biogenesis, is a powerful route to ensure stable and homogeneous presentation of targeting ligands. By genetically engineering donor cells, such as HEK293T cells, MSCs, or DCs, the desired ligand is commonly fused to an endogenous exosomal anchor, like CD63 or CD81. This approach has yielded examples, such as the fusion of the RVG peptide to the exosomal protein Lamp2b [180, 181], demonstrating targeted delivery across the BBB and to neuronal cells in vivo. Similarly, the display of specific single-chain variable fragments (scFv) targeting EGFR (EGF/GE11) [166] or tumor-homing peptides like iRGD ensures active recognition and improved uptake by target cells [191]. However, this stability comes at the cost of complexity, requiring the development of stable producer cell lines and robust quality control over cellular changes. Beyond the passive display of targeting ligands, protein engineering of exosomes has recently emerged as a more active and modular strategy that simultaneously addresses cargo loading, pharmacokinetics, and targeting specificity. By rationally fusing functional protein domains to exosomal scaffold proteins, most commonly the tetraspanins CD9, CD63, and CD81, as well as Lamp2b and syndecan-derived anchors, researchers can tailor exosome surface composition with sequence-level precision. The group of EL Andaloussi and colleagues pioneered this direction, demonstrating that engineered tetraspanin scaffolds can be used to display targeting moieties or to recruit RNA-binding proteins for selective endogenous loading of mRNA and ASOs into exosomes during biogenesis [192]. More recently, the same group introduced a VSV-G fusogenic protein engineering strategy that markedly enhances cytosolic cargo release in recipient cells, addressing the long-standing endosomal escape bottleneck of natural EVs [193, 194]. In parallel, Vader and colleagues have systematically dissected how surface protein engineering shapes exosome biodistribution and circulation half-life, showing that fusion of CD47 (“don’t-eat-me” signal) or albumin-binding domains to exosomal scaffolds can substantially reduce mononuclear phagocyte clearance and prolong systemic exposure, an engineering route to PK improvement that does not rely on synthetic PEGylation [167]. Collectively, these protein-engineering strategies convert exosomes from passively secreted vesicles into rationally designed, multifunctional carriers in which loading, circulation, and targeting can be encoded at the genetic level. This integrated engineering paradigm is now being translated by clinical-stage developers such as Evox Therapeutics and Codiak BioSciences, marking a shift from descriptive isolation toward systematic exosome design. Given the inherent complexities of cell engineering, an alternative and highly modular approach lies in the direct modification of the purified exosome. This avoids the need for cell line development and offers versatility through chemical means. Among the various methods for cargo loading, electroporation remains a widely employed and highly effective technique. By momentarily increasing the permeability of the exosome lipid bilayer through electrical pulses, this method facilitates the robust incorporation of both hydrophilic and hydrophobic payloads into the exosomal lumen [195, 196]. Other high efficiency techniques like Click Chemistry (Cu(I)-catalyzed alkyne-azide cycloaddition, Strain-promoted azide–alkyne cycloaddition) are frequently exploited to attach ligands to the exosome surface, providing rapid customization. Wang et al. integrated click chemistry into exosomes, constructing RGD-modified exosomes loaded with the azide sugar precursor Ac4ManNAz. Upon delivery to endothelial cells, these exosomes mediated metabolic labeling of cell-surface glycans, enabling precise, real-time imaging of angiogenesis while simultaneously promoting vascular regeneration both in vitro and in vivo [197]. Furthermore, lipid insertions are crucial not only for active targeting but also for enhancing pharmacokinetics. The insertion of DSPE-PEG lipids, for instance, significantly extends the systemic circulation time by conferring a “stealth” effect against clearance by the macrophage and monocytes. Compared with the control group, the DSPE-PEG-exosomes can be internalized by target cells more effectively and thus, increase the miRNA expression [198]. Nonetheless, these direct modification techniques also face a dilemma. They may result in heterogeneous ligand density, potential damage to the exosome structure during harsh chemical steps and could increase immunogenicity. Hybrids vectors Despite the significant progress achieved by individual non-viral delivery platforms such as LNP, polyplexes, liposomes, and exosomes, each system still faces intrinsic limitations. Recent advances therefore emphasize combinational strategies, which integrate the complementary strengths of different vectors to achieve enhanced delivery efficiency, capacity, biocompatibility, and precision. Such synergistic systems have opened new fields for gene therapy and are rapidly expanding their applications. Exosome-liposome nanovesicles Hybrid exosome–liposome nanovesicles integrate the natural targeting and biocompatibility of exosomes with the structural tunability and high drug loading capacity of liposomes [199, 200]. Such hybrids exhibit enhanced stability and circulation, enlarge the loading capacity, enabling superior therapeutic delivery across biological barriers such as the BBB [201, 202]. An extracellular vesicle–liposome hybrid nanoparticle (EV–LNP) was raised which shows rational hybridization by integrating biological functionality and synthetic precision. By fusing tumor cell-derived EVs with cationic liposomes, the study successfully elevated the targeting specificity and enlarged the loading capacity [203]. Exosomes effectively compensate for the limited biocompatibility of liposomes, while liposomes, in turn, help exosomes enhance their loading capacity. The hybrid system can also be applied in artificially engineered cells. A hybrid system composed of chimeric antigen receptor-T (CAR-T) cell-derived exosome with lung-targeted liposome containing paclitaxel (PTX) called Lip-CExo@PTX. The bispecific CAR-T cells targeting mesothelin (MSLN) and programmed death ligand-1(PD-L1). Thus, the Lip-CExo@PTX can accumulate in the lung tissue due to the lung-targeted liposome, specifically target MSLN-positive tumor cells and synergistically induce immunological death of cancer cells by PD-L1 and PTX. In CT26 metastatic lung cancer model, Lip-CExo@PTX inhibited the growth of tumor and prolonged the survival time (80% of the mice in Lip-CExo@PTX survived and in other groups all the mice died) which verified the efficiency of Lip-CExo@PEX in vivo [204]. The “CAR-exosome-liposome hybrids” have particular significance for future cancer immunotherapy, suggesting a feasible route toward cell-free CAR therapeutics with organ specificity and high targeting. Due to the penetration of BBB, exosomes are also widely used as vectors for neurological diseases, such as Glioma, Alzheimer’s disease and Parkinson’s disease [181, 182, 205]. tLyp1 has affinity to neuropilin receptor1. By linking tLyp1 and tranferrin(Tf) with liposome and blood exosomes, the hybrid system is capable of penetrating BBB and targeting glioma, With the highest fluorescence in brain compared to other groups, the hybrid systems showed enhanced capability to penetrate BBB and target the brain parenchyma. Co-loaded with salvianolic acid B for anti-angiogenesis and cryptotanshinone for cytotoxicity, the pHybrid/SAB-CPT greatly inhibited tumor growth and prolonged survival time in mice [201]. Lipid-polymer hybrid nanoparticles (LPHNPs) As for Lipid-polymer hybrid nanoparticles (LPHNPs), it’s one of the best examples. LPHNPs are typically composed of a polymeric core, a monolayer of lipids, and an outer PEGylated lipid shell [206]. The combination of lipid and polymer enables it stable, high bio-compatibility, increased bioavailability and improved drug entrapment efficacy with controlled release, specific targeting ability [207]. Some antibodies, peptides, aptamers and small molecules may be coated outside the PEGylated lipid to attain active targeting, for example, the folate receptors [208, 209], transferrin receptors [210] and Cluster-of-Differentiation [211]. The incorporation of additional surface receptors endows LPHNPs with enhanced targeting specificity. In an in vivo rat model of metastatic prostate cancer, core–shell LPHNPs co-loaded with docetaxel and a molecularly targeted agent demonstrated superior therapeutic performance, showing significantly reduced systemic toxicity compared with the combination of free FTY720 and docetaxel which led to approximately 20% body weight loss. Notably, researchers further reported that the co-encapsulation of the sphingosine kinase 1 (SK1) inhibitor FTY720 (fingolimod) and docetaxel within LPHNPs markedly mitigated off-target toxicity and, importantly, prevented FTY720-induced lymphopenia which is a major limitation of its systemic administration [212]. Combining LPHNPs with photothermal therapy represents a promising approach for synergistic cancer treatment. In a representative design, dopamine-modified PEI was conjugated with lipid components to form Lip@PSD-siP, a hybrid nanocarrier co-loaded with PD-L1 siRNA and the photosensitizer IR780. The dopamine modification reduced the intrinsic cytotoxicity of PEI and introduced a redox-sensitive disulfide linker, enabling glutathione (GSH) responsive siRNA release in the TME. Given the elevated intracellular GSH concentration in tumor cells [213, 214], the disulfide bonds were cleaved, promoting efficient siRNA liberation, enhanced cellular uptake by CT26 cells, and effective endosomal escape. Consequently, the hybrid system achieved excellent PD-L1 silencing (the PD-L1 positive rate decreased by 82.4%), resulting in inhibition of both primary and distant tumors as well as delayed metastatic progression. This multifunctional platform triggered DC maturation, increased memory T cell populations, and reduced Treg cells, thereby reshaping the tumor immune microenvironment toward an immunostimulatory phenotype. In this hybrid system, the polymer core functions as a responsive genetic carrier that condenses and releases siRNA in a GSH-dependent manner, while the lipid shell provides systemic stability and serves as a photothermal conversion matrix. Through the integration of immune gene silencing, mild photothermal stimulation, and site-specific release, the LPHNP system enables achievable immunological therapy for colorectal cancer [215]. Beyond the conventional core–shell architecture, recent advances have further expanded the structural design of LPHNPs. In particular, a lipid–polymer hybrid “particle-in-particle” nanostructure has been reported [216]. Lipid-modified poly(β-amino esters) (L-PBAEs) condense nucleic acids into nanoscale complexes that are subsequently embedded within a PLGA–PEG matrix. The structure that small particles (L-PBASEs with nucleic acids) in large particles (PLGA-PEG matrix) is called “Particle-in-Particle”. With higher transfection efficiency and stably slow release of nucleic acids, the “Particle-in-Particle” (PNP) construction shows its unique advantages in vaccines. Most importantly, the platform keeps long-term stability after lyophilization for 12 months. By encapsulating plasmid DNA or mRNA encoding the SARS-CoV-2 spike protein, robust spike-specific humoral immune responses are elicited. Importantly, these immunogenic effects were preserved after lyophilization at − 20 °C for 1 year. The excellent transfection efficiency and stability of these nanoparticles can be attributed to two key aspects. First, the design integrates the membrane fusion properties of lipids with the structural robustness and lyophilization compatibility of polymers. Second, the unique particle-in-particle architecture spatially segregates nucleic acid condensation from structural stabilization and release control, thereby optimizing the deficiency of conventional core–shell architecture. Beyond PNPs, this hybrid architecture is being actively leveraged for in vivo CAR-T cell therapy, a burgeoning field where T-cells are genetically modified directly within the patient to bypass costly ex vivo manufacturing [217]. This is achieved by functionalizing the lipid shell with targeting moieties like CD3 or CD5 antibodies, which direct the mRNA-loaded particles to specific immune subsets. These hybrid platforms operate by decoupling functional modules within a single carrier system: polymers primarily condense nucleic acids and provide structural integrity, while lipid layers facilitate membrane interaction, endosomal escape, and systemic transport. This division of labor allows optimization of pharmacokinetics and intracellular trafficking. Furthermore, recent studies on LPHNPs demonstrate enhanced serum stability and sustained release profiles, which are critical for maintaining therapeutic gene expression in vivo. By combining the mechanical protection of the polymeric fraction with the targeting precision of lipid-anchored ligands, LPHNPs have evolved into programmable tools for global genetic immunotherapy. The charm of hybrid systems lies precisely in their ability to combine different materials to create entirely new carrier platforms with excellent multifunctional performance. The exosome–liposome hybrid system and the lipid–polymer hybrid system represents two major paradigms among the diverse hybrid nanocarrier platforms developed to optimize non-viral delivery. Except for these representative examples, the combination of non-viral vectors with various functional materials has also created various hybrid delivery strategies. By integrating gels, nanoceria and metals with the vectors in a modular manner, researchers can finely tune physicochemical properties and further influence the targeting specificity and intracellular responsiveness [114, 218, 219]. Through such rational combinations, hybrid systems overcome the intrinsic limitations of single component carriers, achieving more comprehensive and multi-functional vesicles. Although LPHNPs and exosome-liposome nanovesicles indeed harness the stability and controlled-release properties of synthetic polymers/liposomes together with the biocompatibility and targeting capabilities of biological lipids and exosomes, these hybrid platforms also introduce notable challenges. Specifically, LPHNPs often suffer from difficulties in optimizing the lipid-to-polymer ratio, leading to potential phase separation, drug leakage during storage, and limited encapsulation efficiency for hydrophilic drugs [220–223]. Additionally, their in vivo toxicity, biodistribution, and long-term safety require extensive evaluation. In contrast, exosome-liposome nanovesicles frequently exhibit inconsistent fusion efficiency (e.g., via sonication or extrusion), resulting in heterogeneous vesicle populations, possible damage to exosomal bioactive components [224], source-dependent in vivo distribution [225], rapid clearance [226], and increased difficulties in characterization and purification [224, 227] due to their hybrid biological-synthetic nature [228]. These issues, alongside heightened manufacturing complexity and scalability constraints for both platforms, highlight the need for advanced process optimization and rigorous quality control to facilitate clinical translation of hybrid systems. Other non-viral vectors Beyond the most established platforms, the landscape of non-viral delivery is further enriched by several specialized systems that offer unique advantages. Inorganic nanoparticles, such as gold [229–231], mesoporous silica [232, 233], or magnetic iron oxide [229, 234, 235], provide a level of structural rigidity and surface tunability that organic carriers often lack. Their ability to integrate diagnostic imaging with therapeutic delivery, a field known as theranostics, makes them particularly valuable for complex diseases [231, 236, 237]. However, despite their stability, these inorganic systems often struggle with efficient cytosolic release and long-term clearance from the body, which currently limits their clinical pace compared to lipid-based systems. Similarly, carbon-based vectors like graphene and carbon nanotubes have shown remarkable potential due to their high surface area and ability to penetrate cellular membranes directly [238, 239]. However, the concerns regarding their long-term biocompatibility remain a significant hurdle. In the pursuit of chemical precision, peptide-based carriers offer highly defined architectures that allow for near-atomic control over the delivery process. Peptide-based systems, such as cell-penetrating peptides (CPPs), act as high-efficiency “molecular keys” that can unlock cellular membranes [240, 241]. While they commonly serve as functional components, self-assembling peptide nanofibers allow them to function as primary delivery systems due to their enhanced stability in physiological environments. Perhaps the most cutting-edge addition to non-viral vectors is DNA origami or DNA-based nanostructures. By utilizing the predictable base-pairing of DNA itself to build three-dimensional cages, researchers can create delivery vehicles with extreme precision, allowing for the exact placement of drugs and targeting molecules [242–244]. Although these alternative platforms, ranging from inorganic crystals to programmable DNA shapes, are still navigating the complexities of clinical maturation and large-scale manufacturing, they offer valuable alternative strategies for the field. Overview and comparison of non-viral vectors To help orient design choices, we organize their comparative strengths along four practical dimensions: cargo type, target organ, immunological requirement, and translational maturity (Table 2). Table 2The non-viral vectors under clinical trialsVectorsIndications Delivery systemPhaseClinic trial identifierLNPSolid tumorLNP–mRNA (mRNA-2416, Moderna)Phase 1NCT03323398LNP–mRNA (BNT122, BioNTech)Phase 1/2NCT03289962, NCT03815058, NCT04486378Solid tumor (Complete resection)LNP–mRNA (mRNA-4157, Moderna)Phase 1 NCT03313778 KRAS-mutated solid tumorLNP–mRNA (mRNA-5671, Moderna)Phase 1NCT03948763CLDN6-positive solid tumorLNP–mRNA (BNT211, BioNTech)Phase 1/2NCT04503278MelanomaLNP–mRNA (mRNA-4157, Moderna)Phase 3 (KEYNOTE-942)NCT03897881LNP–mRNA (BNT111, BioNTech)Phase 2NCT04526899SARS-CoV-2 (COVID-19)mRNA (BNT162b1, BioNTech/Pfizer)Phase 1NCT04523571LNP–mRNA (ChulaCov19, non-stabilized spike)Phase 1NCT04566276LNP–mRNA (COVID-19 vaccine)Phase 1NCT05477186LNP–self-amplifying mRNAPhase 1NCT04758962LNP–mRNAPhase 1NCT04798027LNP–self-amplifying mRNA (VLPCOV01)Phase 1jRCT2051210164LNP–mRNA (BNT162b1/BNT162b2a)Phase 1–3NCT04368728LNP–mRNA (HERALD study)Phase 2b/3NCT04652102LNP–mRNA (mRNA-1273, Moderna) aPhase 3NCT04470427LNP–mRNA (DS-5670d, Omicron XBB.1.5)Phase 3jRCT2031230424LNP–mRNA (COVID-19 vaccine)Phase 3NCT04480957RabiesLNP–mRNA (Rabies vaccine)Phase 1NCT03713086Hereditary transthyretin-mediated amyloidosis (hATTR)LNP–CRISPR/Cas9 (NTLA-2001)Phase 1NCT04601051LNP–siRNA (patisiran, targeting TTR)bPhase 3NCT01960348Hepatic fibrosisLNP–siRNA (BMS-986263, targeting HSP47)Phase 1NCT03142165LNP–siRNA (BMS-986263)Phase 2NCT03420768H10N8 and H7N9 Influenza virusesLNP–mRNAPhase 1NCT03076385Glioblastoma multiforme (GBM)LNP–mRNA (CVGBM mRNA vaccine)Phase 1NCT05938387Ebola virus diseaseLNP–siRNA (TKM-130803)Phase 2PACTR201501000997429Heterozygous familial hypercholesterolemia (HeFH), atherosclerotic cardiovascular disease (ASCVD), and uncontrolled hypercholesterolemiaLNP-CRISPR/Cas9 (VERVE-101, targeting PCSK9)Phase 1bNCT05398029Respiratory Syncytial Virus (RSV)LNP–mRNA (mRNA-1345, Moderna)Phase 1–2NCT04528719Auto-immunological alterationLNP–mRNA6231 (IL-2 cytokine)Phase 1NCT04916431PolyplexesPancreatic ductal adenocarcinoma a complex of pDNA and linear PEI (jetPEI 22 kDa) in combination with gemcitabine(CYL-02)Phase2NCT02806687Pancreatic cancer siG12D-LODERphase1/2 NCT01676259Advanced ovarian, fallopian tube or primary peritoneal cancersPEG-PEI-Cholesterol-pDNAphase2 NCT05739981NCT03393884Non-melanoma skin cancersjet-PEI pDNAPhase 1bNCT04160065Soft tissue sacromajetPEI dsRNAPhase 1NCT04420975Unresectable malignant melanomajetPEI dsRNAPhase 2NCT04570332Metastatic refractory non-small cell lung carcinomajetPEI dsRNAPhase1b/2NCT05265650NeuroblastomaPEI DNAPhase 1NCT04049864Advanced stage solid tumorsPEI-PEG TAT peptide-pDNAPhase 1NCT05578820LipoplexesMelanomamRNA-lipoplexPhase 1/2NCT02410733PDACmRNA-lipoplexPhase 1NCT04161755PDACsiRNA-lipoplexPhase 1NCT00938574PDACsiRNA-lipoplexPhase 1b/2aNCT01808638ExosomesDysregulation of NF-κB activationengineered, allogeneic EV derived from HEK293 cells and loaded with NF-kB (ILB-202)Phase 1NCT05843799PDACEngineered exosomes with KrasG12D specific siRNAPhase1bNCT03608631NSCLCDendritic cell–derived exosomesPhase 2NCT01159288a Reflects formulations that have received full approval from FDA and conditional marketing authorisation from EMAb Reflects formulations that have received full regulatory approval from both FDA and EMA The non-viral vectors under clinical trials a Reflects formulations that have received full approval from FDA and conditional marketing authorisation from EMA b Reflects formulations that have received full regulatory approval from both FDA and EMA Table 1The applications, advantages and challenges of non-viral vectorsDelivery systemApplicationsAdvantagesChallengesLNP• Patisiran ONPATTRO® for hereditary transthyretin amyloidosis• COVID-19 mRNA vaccines (Pfizer/BioNTech BNT162b2, Moderna mRNA-1273)• Cancer vaccinesPersonalized neoantigen mRNA vaccinesmRNA vaccines encoding tumor antigens or immune modulators• Infectious diseasesInfluenza, RSV, CMV, HIV• Hepatic gene silencing/expression modulation• Clinically validated• High encapsulation for mRNA/siRNA• Low toxicity• Scalable and reproducible manufacturing• Good stability and prolonged circulation• Broad therapeutic applicability (vaccines, protein replacement, gene editing)• Inefficient endosomal escape• Anti-PEG immune responses after repeated dosing• Accelerated blood clearance (ABC)• Complement activation–related pseudoallergy (CARPA)• Induction of anti-PEG antibodies• Liver tropism limits targeting flexibilityPolyplexes• Genetic diseases: DMD, CF• Cancer-related gene therapy • siRNA/miRNA silencing • pDNA delivery• Relatively high efficiency• Highly tunable and easy to chemically modify• Large loading capacity• Low production cost and simple preparation• Stimuli-responsive release• Toxicity• Poor hemocompatibility and tendency to aggregate• Weak serum stability• Limited clinical translation Lipoplexes•Local or regional delivery for PDAC• High transfection efficiency, especially in vitro• Well-established lipid formulation technologies• Higher toxicity compared to LNPs• Poor serum stability• Most clinical applications have been replaced by modern LNP systemsExosomes•Cancer gene therapy•Immunotherapy (DEX vaccines)•Neurological therapies• Natural origin with excellent biocompatibility• Low immunogenicity• Intrinsic targeting capability depending on the parent cell type• High stability• Natural fusion with cell membranes facilitates endosomal escape• Low capacity of cargoes• Purification is difficult• Large-scale manufacturing• Unclear mechanisms• Low loading efficiencyLipid-polymer Hybrid Nanoparticles• DNA and mRNA COVID vaccines• Gene therapy (siRNA, mRNA, pDNA delivery)• Drug delivery• Co-delivery of nucleic acids and drugs• Gene editing delivery (CRISPR/Cas systems)• Combine the structural stability of polymer with the biocompatibility of lipids• Improved protection of nucleic acids and enhanced serum stability• Higher transfection efficiency but lower toxicity• Tunable physicochemical properties (size, charge, release)• Co-delivery of multiple therapeutic cargo (drugs, RNAs)• Complex formulation and manufacturing processes• Batch-to-batch variability• Limited large-scale manufacture• Potential toxicity from polymer components• Regulatory pathways not yet well establishedLiposome-exosome Hybrid Nanoparticles • Cancer Immunochemotherapy• Gene therapy• Gene editing delivery (CRISPR/Cas systems)• Drug delivery• Combine the high loading capacity of liposomes with the natural targeting properties of exosomes• Improved delivery efficiency and cellular uptake• Enhanced stability and reduced immunogenicity• Hybridization methods are still not standardized• Difficult quality control and characterization• Large-scale production remains challenging• Potential heterogeneity in hybrid vesicles• Mechanisms of targeting and intracellular trafficking remain unclear The applications, advantages and challenges of non-viral vectors • Patisiran ONPATTRO® for hereditary transthyretin amyloidosis • COVID-19 mRNA vaccines (Pfizer/BioNTech BNT162b2, Moderna mRNA-1273) • Cancer vaccinesPersonalized neoantigen mRNA vaccinesmRNA vaccines encoding tumor antigens or immune modulators • Infectious diseasesInfluenza, RSV, CMV, HIV • Hepatic gene silencing/expression modulation • Clinically validated • High encapsulation for mRNA/siRNA • Low toxicity • Scalable and reproducible manufacturing • Good stability and prolonged circulation • Broad therapeutic applicability (vaccines, protein replacement, gene editing) • Inefficient endosomal escape • Anti-PEG immune responses after repeated dosing • Accelerated blood clearance (ABC) • Complement activation–related pseudoallergy (CARPA) • Induction of anti-PEG antibodies • Liver tropism limits targeting flexibility • Genetic diseases: DMD, CF • Cancer-related gene therapy • siRNA/miRNA silencing • pDNA delivery • Relatively high efficiency • Highly tunable and easy to chemically modify • Large loading capacity • Low production cost and simple preparation • Stimuli-responsive release • Toxicity • Poor hemocompatibility and tendency to aggregate • Weak serum stability • Limited clinical translation • High transfection efficiency, especially in vitro • Well-established lipid formulation technologies • Higher toxicity compared to LNPs • Poor serum stability • Most clinical applications have been replaced by modern LNP systems •Cancer gene therapy •Immunotherapy (DEX vaccines) •Neurological therapies • Natural origin with excellent biocompatibility • Low immunogenicity • Intrinsic targeting capability depending on the parent cell type • High stability • Natural fusion with cell membranes facilitates endosomal escape • Low capacity of cargoes • Purification is difficult • Large-scale manufacturing • Unclear mechanisms • Low loading efficiency • DNA and mRNA COVID vaccines • Gene therapy (siRNA, mRNA, pDNA delivery) • Drug delivery • Co-delivery of nucleic acids and drugs • Gene editing delivery (CRISPR/Cas systems) • Combine the structural stability of polymer with the biocompatibility of lipids • Improved protection of nucleic acids and enhanced serum stability • Higher transfection efficiency but lower toxicity • Tunable physicochemical properties (size, charge, release) • Co-delivery of multiple therapeutic cargo (drugs, RNAs) • Complex formulation and manufacturing processes • Batch-to-batch variability • Limited large-scale manufacture • Potential toxicity from polymer components • Regulatory pathways not yet well established • Cancer Immunochemotherapy • Gene therapy • Gene editing delivery (CRISPR/Cas systems) • Drug delivery • Combine the high loading capacity of liposomes with the natural targeting properties of exosomes • Improved delivery efficiency and cellular uptake • Enhanced stability and reduced immunogenicity • Hybridization methods are still not standardized • Difficult quality control and characterization • Large-scale production remains challenging • Potential heterogeneity in hybrid vesicles • Mechanisms of targeting and intracellular trafficking remain unclear The physicochemical properties of the nucleic acid cargo largely dictate which vector class is suitable. For mRNA and self-amplifying RNA, LNPs remain the benchmark, owing to high encapsulation efficiency conferred by ionizable lipids and to clinically validated performance in the BNT162b2 and mRNA-1273 vaccines [23, 24]. For small, structurally stable oligonucleotides such as siRNA and ASOs, both polyplexes and lipoplexes are feasible, but polyplexes provide an additional degree of freedom: stimuli-responsive backbones (e.g., redox-, pH-, or esterase-cleavable linkers) enable triggered intracellular release that lipid systems do not readily support [39, 70, 76]. For plasmid DNA, where cargo size and the requirement for nuclear entry are limiting, branched architectures such as PEI and PBAE accommodate larger payloads more efficiently than LNPs, whose packaging capacity is intrinsically constrained [63, 67, 93]. For gene-editing payloads, LNPs have already been validated clinically through NTLA-2001, which delivers CRISPR-Cas9 components targeting TTR in hereditary transthyretin amyloidosis (NCT04601051), although hybrid platforms are increasingly explored when extra-hepatic editing is required [60]. For endogenous biomolecules such as miRNAs and functional proteins, exosomes are the natural choice, as their biogenesis pathway selectively packages such cargoes and preserves their bioactivity during transfer to recipient cells [119, 120]. Tropism determines whether a delivery strategy is even feasible for a given indication. Following systemic intravenous administration, conventional LNPs accumulate predominantly in the liver through ApoE-mediated uptake; this default behavior is exploited in patisiran for hepatic TTR silencing [37]and is therefore an asset rather than a liability when the liver is the intended target. For extra-hepatic tissues, two engineering strategies have redirected biodistribution: SORT-LNPs achieve spleen- and lung-selective mRNA delivery by tuning the proportion of charged helper lipids [58], whereas the more recent POST platform encodes organ tropism via short peptide motifs displayed on a PEGylated sheath, achieving delivery to lung, spleen, placenta, and bone marrow without relying on cationic charge [60]. Polyplexes provide complementary access to inflamed tissue, as illustrated by the guanidinium-rich PONI-Guan polymer that accumulates in inflamed lung [68]. For pulmonary delivery via inhalation, particularly relevant in cystic fibrosis and lung cancer, polyplexes built from lyophilization-tolerant polymers such as PEI and HPAE3 retain transfection competence after spray drying and aerosolization, an advantage not yet matched by lipid systems [92, 93]. For CNS delivery, where the blood–brain barrier excludes most synthetic carriers, exosomes (and exosome–liposome hybrids) display the most robust evidence: RVG-Lamp2b-engineered exosomes deliver siRNA and mRNA to neurons in models of Alzheimer’s and Parkinson’s disease [135–137], and tLyp1/transferrin-decorated exosome–liposome hybrids further improve glioma penetration relative to either parent vector [150] (see Exosome-Liposome nanovesicles section). For local or intratumoral delivery, lipoplexes and LPHNPs provide adequate stability without requiring the stealth modifications needed for systemic circulation, as demonstrated by the autogene cevumeran (BNT122) regimen in pancreatic cancer [106, 109] and docetaxel/FTY720 co-loaded LPHNPs in metastatic prostate cancer [159]. Because non-viral carriers themselves engage the innate immune system, the immune profile of a vector should be matched to the therapeutic intent rather than treated as a uniform safety concern. When immune activation is desired, as in prophylactic and therapeutic vaccination, the intrinsic adjuvanticity of ionizable LNPs and cationic lipoplexes becomes a feature: both have advanced to large pivotal trials—LNPs in COVID-19 mRNA vaccines and lipoplexes in personalized neoantigen vaccines for pancreatic and other cancers. When immune evasion is required, such as in chronic gene correction, repeated dosing, or delivery to immunologically sensitive tissues, exosomes offer a clear advantage owing to native surface markers (e.g., CD47) that suppress phagocytic clearance and to overall low immunogenicity in autologous settings. Surface engineering of synthetic carriers can partially close this gap: zwitterionic shielding (e.g., PONI-Guan-Zwit) reduces opsonization while preserving siRNA delivery [49], and PEG- or polysaccharide-shielded polyplexes mitigate complement activation although anti-PEG antibodies and accelerated blood clearance remain unresolved limitations even for clinically deployed LNPs [41, 44, 45]. When immune remodeling of the tumor microenvironment is the objective, polyplexes provide an attractive vehicle because they can co-condense large immunomodulatory payloads such as IL-12 plasmids [39] or CRISPR/dCas9-KRAB constructs targeting PD-L1 [76], and LPHNPs further integrate such gene cargoes with photothermal or chemotherapeutic components for synergistic effects [165]. Even an optimal mechanistic match is insufficient without compatible manufacturing and a regulatory pathway. LNPs are the most translationally mature platform, with established GMP production, FDA-approved products (Onpattro, Comirnaty, Spikevax) and over 200 ongoing RNA-based clinical trials; in the absence of a specific reason to choose otherwise, LNPs are the default starting point. Lipoplexes and polyplexes occupy an intermediate position, with several formulations in mid- to late-stage clinical evaluation, including autogene cevumeran (BNT122/RO7198457; NCT02410733, NCT05968326) for pancreatic cancer [106, 109] and the jetPEI-based CYL-02 in pancreatic ductal adenocarcinoma (NCT02806687); polyplex chemistry remains highly tunable, but batch-to-batch variability and the absence of consensus characterization standards continue to constrain large-scale manufacturing. Exosomes and hybrid systems are the least mature: although engineered exosomes such as ILB-202 (NCT05843799) and KrasG12D-targeting exosomes (NCT03608631) have entered phase 1 trials, the field still lacks standardized purification, scalable GMP production, and quality-control assays that distinguish exosomes from co-isolated extracellular vesicles [141–143]. Hybrid platforms inherit the manufacturing complexity of both parent systems and at present are largely confined to preclinical proof-of-concept [152]. Clinical-stage examples for each vector class are compiled in Table 2. Conclusion The past decade has reframed non-viral gene delivery from a search for the single best vector to a structure–function design problem in which therapeutic context dictates carrier choice. Across the four vector classes and the hybrid architectures reviewed here, three converging design principles can be distilled. Endosomal escape, rather than cellular uptake, has become the rate-limiting step for nearly all non-viral platforms. This recognition has driven the universal adoption of pH-titratable chemistry, ionizable lipids in LNPs [5, 7], proton-buffering amines in PEI- and PBAE-based polyplexes [58, 81], and fusogenic peptides such as GALA and INF7 in engineered exosomes, as the shared mechanistic solution across otherwise dissimilar materials. Moreover, organ tropism is increasingly treated as a programmable property rather than an intrinsic vector trait. The progression from ApoE-mediated hepatic accumulation of conventional LNPs [245] to charge-tuned SORT formulations [80], peptide-encoded POST particles [82], and ligand-engineered exosomes (RVG-Lamp2b [180], GE11, iRGD [88, 197]) reflects a conceptual shift from passive biodistribution toward rationally specified delivery. Immune compatibility has shifted from a passive safety concern to an active design lever: certain vectors are deliberately engineered to provoke immunity, as in mRNA-LNP and lipoplex vaccines, whereas others are shielded with PEG, zwitterionic, or biological coatings to evade clearance during chronic gene correction. These principles, however, cannot be simultaneously optimized within any single material class. Lipid systems excel in mRNA encapsulation and clinical scalability but remain constrained by hepatic tropism and anti-PEG immunity. Synthetic polymers offer modular tunability and large cargo capacity but pay for these advantages in cytotoxicity and serum instability. Lipoplexes provide simple, robust formulations for vaccine and local applications but have been largely superseded by ionizable LNPs in systemic settings. Exosomes deliver native biocompatibility and intrinsic tropism, particularly across the blood–brain barrier [135–137], but remain limited by low loading capacity, vesicle heterogeneity, and the absence of standardized large-scale manufacturing [141]. The emergence of hybrid architectures, exosome–liposome [152, 153], lipid–polymer [221], and particle-in-particle [216] constructs, is not a stylistic trend but a structural necessity. Hybrids are the configurations in which complementary properties can coexist, allowing escape efficiency, organ tropism, and immune compatibility to be tuned along independent axes rather than traded against one another. This rationale also clarifies why hybridization has progressed most rapidly in indications that no single vector adequately serves: glioma and other CNS tumors, where BBB penetration must coexist with cytotoxic payload delivery [150]; lung-targeted CAR-mimetic immunotherapy, where organ specificity must be combined with antigen recognition [153]; and lyophilizable nucleic acid vaccines, where transfection competence must survive long-term storage [166]. A genuine bench-to-bedside translation, however, introduces a tension that the field has not yet resolved. The compositional complexity that gives hybrid systems their performance advantage is in direct conflict with regulatory standardization, characterization reproducibility, and GMP scalability, the very attributes that enabled LNPs to become first-in-class. Future directions With the rise of artificial intelligence (AI), the creation of non-viral vectors and hybrid systems is going to undergo a profound transformation. Leveraging machine learning, neural networks, expert systems and deep learning [246, 247], AI can digitize measurable biological features and learn complex non-linear relationships from large, massive data and experiences, predict optimal design in the gene transduction procedures, from unspecific interactions with serum components to cellular uptake, endosomal escape and nuclear entry. A predictive model of silver nanoparticles protein corona to minimize unspecific combination in serum using random forest classification was established. Taking nanoparticles’ properties, protein characteristics and interaction conditions into consideration, the model found the correlation between the variables with protein corona population and produced a weighted list of features involved in the interaction [248]. With the help of ML, the cost, labor and time consumed for satisfying results is decreased significantly. And that’s why the summary of past successful non-viral vectors is necessary. The integration of AI is the first step toward a broader transformation where non-viral and hybrid vectors evolve into “environment-aware” systems. While AI helps us predict design outcomes, the most promising future for these platforms lies in programmed autonomy, creating vectors that do not just passively carry cargo but actively sense and respond to biological cues [249, 250]. Instead of relying on static surface ligands, the next generation of delivery tools will likely utilize “logic-gate” responses, undergoing structural changes only when they encounter specific triggers like the acidic pH of a tumor or unique intracellular enzymes [251]. To move past the persistent bottleneck of endosomal entrapment, the field is shifting away from simple chemical interactions toward viral-mimetic fusion strategies. The most effective road forward involves incorporating “molecular drills” or pH-sensitive fusogenic lipids that can actively disrupt the endosomal membrane, mimicking the high-efficiency escape mechanisms used by viruses [236, 252]. By integrating these bio-inspired motifs into hybrid polymer-lipid systems, we can transition from passive entrapment to near-total cytosolic delivery. This biological sophistication must also extend to how we handle the bloodstream. A transformative shift is occurring where PEG is no longer used to simply “hide” from the immune system, but instead engage in active corona engineering. By designing surfaces that intentionally recruit specific plasma proteins, such as Apolipoprotein E, we can turn the body’s natural response into a targeting advantage, using the recruited “protein corona” as a guide to reach specific organs or cross the BBB [245, 253, 254]. Beyond chemical composition, the mechanical properties of the vector, such as its stiffness and shape, represent an overlooked design frontier that will play a crucial role in how deeply these particles can penetrate dense tissues like solid tumors. However, these advanced designs will only fulfill their therapeutic potential if they can bridge the gap between the bench and the bedside. The most viable path toward successful translation is the adoption of a “plug-and-play” modularity. By treating the vector as a standardized, validated “chassis” that can be easily customized with different therapeutic modules, we can simplify the regulatory process and improve reproducibility [255–258]. When coupled with microfluidic manufacturing, which allows for precise and scalable production, these modular platforms will provide the consistency and efficiency required to transform experimental gene therapies into widely accessible clinical treatments [259–261]. Author contributions Concept and design: Zhigong Wei and Wenlin Wu, data collection and analysis: Wenlin Wu, drafting of the article: Wenlin, Wu, critical revision of the article for important intellectual content: Xingchen Peng and Zhigong Wei, study supervision: Xingchen Peng and Zhigong Wei. All the authors approved the final article. Funding The work was supported by the Noncommunicable Chronic Diseases-National Science and Technology Major Project (2023ZD0503000, 2023ZD0503004), the Regional Innovation and Development Joint Fund Key Project of the National Natural Science Foundation of China (U24A20735), the National Natural Sciences Foundation of China (82473434), the Institutional Joint Innovation Fund from Sichuan University and Nuclear Power Institute of China (1-KJ-FWHT-WU-20240150), Sichuan Provincial Science and Technology Department Key Research and Development Program (2022YFSY0012), Sichuan Science and Technology Program (2024YFHZ0041, 2024ZYD0054), Science and Technology Project of Sichuan Provincial Health Commission (Clinical Research Special Project JH2023082), the International Science and Technology Cooperation Program of Chengdu Science and Technology Bureau (2024-YF06-00011-HZ and 2022-GH03-00004-HZ), the Strategic Cooperation Special Fund of Sichuan University-Dazhou Municipal People’s Government (2022CDDZ-16), the Science and Technology Cooperation Special Fund of Sichuan University-Zigong (2021CDZG-24), the Health Research Project of Chengdu Eastern New Area Management Committee (202304), 1.3.5 project for disciplines of excellence from West China Hospital of Sichuan University (ZYYC23006), Clinical Research Incubation Project of West China Hospital (23HXFH001), Yunnan Province Key Laboratory of Precision Diagnosis and Treatment for Thoracic Diseases (202449CE340026) and the Ministry of Education University-Industry Collaborative Education Program (230720523707281), Sichuan Provincial Science and Technology Department Key Research and Development Program (2025YFHZ0151). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Data availability No datasets were generated or analysed during the current study. Declarations Ethics approval and consent to participate Not applicable. Consent for publication All authors have provided consent for the manuscript to be published in Journal of Nanobiotechnology. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Contributor Information Zhigong Wei, Email: weizg10@wchscu.edu.cn. Xingchen Peng, Email: pxx2014@163.com. References Associated Data Data Availability Statement No datasets were generated or analysed during the current study.