Biomaterial-Based Nucleic Acid Delivery Systems for In Situ Tissue Engineering and Regenerative Medicine

Viral vectors such as adenovirus (AV), adeno-associated virus (AAV) and lentivirus (LV) are widely employed in regenerative medicine to deliver therapeutic genes that promote cell regeneration, tissue repair and functional recovery.

AVs have advantages of high efficiency, low pathogenicity and high viral titers. And they do not integrate into the host cell genome in vivo, making them suitable for transient gene expression. These vectors efficiently package DNA fragments into infectious viral particles, which mediate DNA delivery and expression in cells. While adenoviruses are widely used in gene therapy, safety concerns, particularly dose-dependent toxicity and immune response, require careful evaluation before clinical practice.

Adeno-associated virus exhibits lower immunogenicity compared to adenovirus, especially recombinant AAV (rAAV). By removing the Rep and Cap genes and retaining the inverted terminal repeats of the virus, rAAV vectors safely package and carry foreign genes [41]. The genome of rAAV is naturally in the form of single-stranded DNA, which is converted to a double-stranded form by host cell DNA polymerase and subsequently transferred to the nucleus [42]. In the nucleus, rAAV genome entries targeted cells without integration into the host’s genome and it keeps the form of free, thereby minimizing the risk of insertional mutagenesis [43,44]. Also, AAV has a variety of serotypes, each of which has diverse tissue tropisms, providing a range of options and allowing tissue-specific targeting. For example, AAV6 is tropic to nerve cells, while AAV9 is tropic to cardiomyocytes [45,46]. AAV’s ability to cross the blood–brain barrier and transduce to skeletal muscle cells and cardiomyocytes with higher efficiency makes it a promising candidate for gene therapy in regenerative medicine [47].

Lentiviral vectors are gene transfer tools, derived from modified human immunodeficiency virus (HIV) or other lentiviral viruses. They are produced and assembled in packaging cells and released from cell membranes by budding. Once they enter the host cell, the LV RNA would be transcribed into DNA, which integrate transgenes into the host genome via reverse transcriptase and integrase enzymes. Lentiviral vectors can effectively infect both dividing and non-dividing cells, such as neurons and muscle cells. Through genetic engineering, LVs can target-specific cell subpopulations, such as astrocytes, to address localized cellular dysfunction without affecting other cells [48]. Also, the capacity of LV to accommodate relatively large transgenes is better than AAVs. Although the safety of lentiviral vectors has gradually improved, the immunogenicity and unpredictable integration sites remain challenges requiring further optimization.

Non-viral vectors, like exosome, lipid-based systems, peptide-based gene delivery systems, polysaccharide derivatives, and nucleic acid-based nanomaterials, have become a research hotspot due to their low immunogenicity and engineerable design. However, they commonly face challenges such as low transfection efficiency or inadequate stability in vivo, necessitating surface modifications (e.g., PEGylation) or optimization of targeting ligands to balance safety and efficacy.

Exosomes are nanoscale vesicles secreted by cells and characterized as low immunogenicity, high biocompatibility and excellent ability to cross biological barriers. They effectively facilitate nucleic acid delivery to targeted cells and promote regeneration through cell-to-cell communication. Nearly all cell types naturally secrete exosomes, including endothelial cells, MSCs, and immune cells [49]. These vesicles can be classified into two groups: natural exosomes and engineered exosome, according to whether they have been artificially modified. Natural exosomes are preloaded with bioactive cargo such as proteins, glycoconjugates, lipids, and nucleic acids [50]. In comparison, engineered exosomes are modified to deliver specific therapeutic molecules as miRNAs and lncRNAs. They can be produced by transfecting targeted genes into stem cells with viral vectors, such as lentivirus and retrovirus [51,52]. This can also be achieved by transfection with non-viral methods like CaCl2 treatment and CD9-HuR fusion protein systems [53,54]. In order to enhance transfection efficiency, especially for difficult-to-transfect cell types like stem cells, advanced electroporation techniques, such as the cellular nanoelectroporation system based on track-etched membranes, have also been developed. These methods ensure cell viability and avoid cytotoxicity concerns while enabling efficient loading of large molecules like CRISPR systems [55]. Electroporation is particularly well suited for CRISPR systems due to its ability to efficiently deliver large macromolecules. By optimizing parameters such as voltage and pulse duration, the loading efficiency of CRISPR components can be significantly enhanced [56]. Once engineered, exosomes deliver nucleic acids cargo into cells via membrane fusion or endocytosis. For example, exosomes loaded with miR-181b were shown to fuse RAW264.7 macrophages, releasing miR-181b which will be overexpressed and regulates the polarization state of macrophages (Figure 4A). Thus, miR-181b promote M2 macrophage polarization and secretion of anti-inflammatory cytokines such as IL-10 to inhibit the inflammatory response by targeting PRKCD and activating the AKT signaling pathway. At the same time, MiR-181b promotes osteogenic factors like VEGF and BMP-2, which ultimately enhance bone regeneration and implant integration with bone tissue indirectly [51].

Despite their promising potential, exosome applications face the following challenges: (1) limited targeting precision without surface modification or engineering strategies; (2) poorly understood drug release mechanisms, complicating spatiotemporal control; (3) their large-scale production faces difficulties in standardization, and the potential immune responses as well as the safety of tumor targeting need to be verified [59].

Liposomes and lipid nanoparticles (LNPs) have emerged as promising gene delivery platforms, and some have been approved for nucleic acid delivery in clinic due to their low toxicity and no immunogenicity. The difference between liposomes and lipid nanoparticles is mainly in morphology. Structurally, liposomes are lipid bilayer vesicles enclosing a hydrophilic core to embed nucleic acids (Figure 4B), whereas LNPs pack tightly nucleic acids between two lipid monolayers in sandwich formation (Figure 4C) [57].

LNP–nucleic acid complexes can be prepared by microfluidic technology. Lipid components (alcohol phase) and nucleic acid oligomers (aqueous phase) can aggregate in microchannels and self-assemble into stable LNP–nucleic acid complexes [60]. Compared to liposomes, LNPs are thought to be more efficient in siRNA delivery because siRNA and lipid are co-assembled during cellular uptake. LNPs incorporate siRNAs with higher concentrations and significantly enhance siRNA uptake efficiency compared to the liposome delivery system [61]. In addition, LNPs can also be modified by combining with proteins to improve performance. For example, Rabbani et al. introduced supercharged coiled-coil protein (CSP) to the previous cationic lipid nanoparticles (CLNs) to enhance siRNA condensation and improve topical siRNA delivery. Thus, CSP enhances siRNA complexation with a CLN and delivery without compromising liposome elasticity. Moreover, some cationic LNPs have special properties. For example, Yuanfeng Li et al. developed a dual-functional lipid nanocarrier (cLpT@siRNA) that integrates ROS-scavenging lipid LpT (Lipid-pba-Tempo) with cationic lipid DOTAP to condense siMMP9 into ~45 nm stable complexes via electrostatic interactions. The nanocomplexes are internalized by macrophages through macropinocytosis and clathrin-mediated endocytosis, escape endosomes via proton sponge effects, and release siRNA to silence MMP9, while LpT eliminates ROS and polarizes M1 to M2 macrophages. This synergistic modulation alleviates inflammation, preserves the extracellular matrix, and enhances angiogenesis and collagen deposition, leading to significantly accelerated wound closure in diabetic mice [62]. In contrast, liposomes face limitations in systemic delivery of nucleic acids due to reticulo–endothelial system uptake and hepatic clearance [58]. Nevertheless, liposomes remain effective transfection agents due to their straightforward preparation. Commercially available lipid-based reagents are widely used, as they can be directly mixed with nucleic acids to form stable lipoplexes [63,64,65]. In other research, liposomes modified by electrostatic interaction were complexed with pDNA to develop PEGylated SA lipoplexes (pegSA lipoplexes), which increased the residence time of pDNA in the bloodstream and promoted pDNA accumulation in bone marrow, enabling their local delivery and high adhesion [66].

Lipid-based systems, including novel lipidoids and cholesterol derivatives, are versatile non-viral gene delivery platforms for gene therapies in regenerative therapies. However, early cationic liposomes exhibited relatively high toxicity, LNP systems may induce inflammatory responses, necessitating optimization of lipid components. Most studies have focused more on the transfection efficiency of liposomes and LNPs than on their specificity and toxicity [67]. More research is needed in balancing scalability, safety, and adaptability for diverse nucleic acid payloads.

RALA (Ras-like protein Ral-A) is a small GTPase that belongs to the Ras superfamily. Its structure primarily consists of a G-domain and a C-terminal domain. The C-terminal domain of RALA can interact with acidic lipids such as phosphatidylinositol on cell membrane, which helps localize the RALA protein to cell membrane, directing its membrane localization and enabling it to interact with other proteins. Furthermore, RALA contains polar amino acid residues that interact with water (hydrophilic moiety) and non-polar amino acid residues that do not easily interact with water (hydrophobic moiety). This amphipathic fusogenic structure enables RALA self-assembly into nanoparticles (NPs) in aqueous environments and makes it widely recognized for its efficacy in nucleic acid delivery. Chambers et al. firstly demonstrated that RALA can efficiently deliver miR-26a to MSCs in vitro, regulating osteogenic signaling. At an N:P ratio of 6, the RALA/miR-26a complexes can form NPs with a hydrodynamic size of ~74 nm and realize miR-26a complete encapsulation [68]. E. Mulholland et al. developed one RALA with a sequence of “N-WEARLARALARALARHLARALARALRACEA-C” including 11alanines, 10 arginines, and a terminal cysteine residue that facilitates disulfide bond formation or peptide immobilization [69]. This special design protests the cargo associated with RALA from enzymatic degradation, enhanced endosomal escape, and ensures cytoplasmic release [70]. Other modifications such as integrating new fragments, like VGVAPG, can further improve transfection efficiency and serum stability. Research on new designed peptides is also in progress, such as nuclear localization signal peptide (NLS), star-shaped poly(L-lysine) polypeptides (star-PLLs) and cell-penetrating peptide, Golgi-ER transport protein (GET) or heterochromatin protein 1 homolog 1 (Hph-1) [71,72,73]. Nuclear localization signal (NLS) peptides and cell-penetrating peptides (e.g., GET, Hph-1) enable efficient intracellular delivery. Bioinspired star-shaped poly(l-lysine) polypeptides (star-PLL) represent a novel class of non-viral vectors. Advances in N-carboxyanhydride polymerization allow precise synthesis of star-PLLs with controlled molecular weights, the number of attached poly(L-lysine) arms, narrow polydispersity, and tailored architectures while preserving chirality. These polypeptides self-assemble into nano-sized complexes with large quantity [74]. This precision allows researchers to create libraries of star-PLLs with diverse architectural motifs, which can be tailored to match specific cellular uptake preferences. For instance, different cell types may exhibit distinct affinities for specific structural variants of star-PLLs. Additionally, starzPLLs are cost-effective and straightforward to produce. Leveraging their biocompatible peptide side chains, they self-assemble rapidly into nano-sized complexes, capable of encapsulating large nucleic acids payloads. David P. Walsh et al. explored three star-PLL variants and demonstrated their ability to deliver pDNA to MSCs via a rapid clathrin-independent uptake mechanism. This process supports non-taxic transgene expression in vitro [75].

Peptide-based nanomedical platforms hold significant potential for delivering nucleic acids to stem cells in regenerative medicine applications due to their biocompatibility, modular design, and scalable synthesis. Nevertheless, it should be noted that they have similar drawbacks, such as non-specific internalization, fast elimination from the body, intracellular or vesicular entrapment. Several strategies for improving peptide-based gene delivery systems are currently under investigation. The most commonly employed approaches to enhance cellular uptake, selectivity, and stability involve chemical modifications, such as conjugating a stearyl moiety to the N-terminus of acid-activated cell-penetrating peptides or substituting L-amino acids with their D-isomers [67,76].

Chitosan, a polysaccharide derived from crustacean exoskeletons, possesses positively charged amine groups that enable electrostatic binding to negatively charged nucleic acids like DNA. Chitosan exhibits excellent biocompatibility, biodegradability, and promotes cell adhesion and growth [77]. However, its transfection efficiency has historically been suboptimal, particularly in primary cells. In one study, a 0.1% concentration of 18 KD chitosan nanoparticles achieved a transfection rate of 18.43% compared to 40.57% for commercial lipofectamine (p < 0.01) [78]. Transfection efficiency is heavily influenced by molecular weight: high-molecular-weight chitosan is more stable but does not release pDNA efficiently, while low-molecular-weight chitosan enhances pDNA release effectively despite reduced stability [79]. Rosanne M. Raftery et al. investigated two types of chitosan as MSC transfection agents—one is 160 kDa medium-molecular-weight polymeric chitosan, and another is 7.3 kDa low-molecular-weight oligomeric chitosan (OCS). This study has demonstrated that OCS–pDNA nanoparticles at an N/P ratio of 20 carrying 2 μg of pDNA is the most optimal formulation, achieving 45% transfection efficiency in monolayer MSCs without cytotoxicity [80]. Nanoparticle size critically impacts chitosan’s efficacy. Traditional methods like the ionotropic gelation technique or complex coacervation are often employed to yield chitosan–pDNA complexes with polydisperse particles [81,82,83,84]. In contrast, microfluidic systems enable precise control over chitosan nanoparticle formation, producing homogeneous nanoparticles, dense particles with a narrow size distribution and reduced cytotoxicity [85]. By precisely adjusting flow rates, solute concentrations, and channel diameters, researchers can tailor chitosan particle size, shape, and compactness [86].

Hyperbranched polysaccharides, like glycogen and amylopectin, are biocompatible and biodegradable but lack inherent positive charges. To combine with siRNAs, they need to be modified with 1,2-ethylenediamine (EDA) and diethylenetriamine (DETA) to carry positive charges. Biyun Lan et al. modified four types of hyperbranched cationic polysaccharide derivatives (HCP) with glycogen and amylopectin. The HCP/siRNAs complexes safely delivered siRNA via the “proton sponge” effect, facilitating endosomal escape and cytoplasmic release. The result shows that glycogen and amylopectin are both efficient and safe platforms to deliver siRNAs [87]. While their transfection efficiency is influenced by molecular weight and degree of deacetylation, there may be a risk of mild inflammatory responses in vivo. Their utility for other nucleic acids requires further exploration [88].

Nucleic acid-based nanomaterials have emerged as versatile platforms due to their precise structural programmability. Tetrahedral framework nucleic acids (tFNAs) are pyramid-shaped, three-dimensional nanostructures assembled via complementary base pairing of four single-stranded DNA. They exhibit straightforward synthesis, wide applicability, structural stability, self-assembly capability, antioxidant properties and anti-inflammatory effects [89,90]. tFNA are categorized into three generations, with the third generation being the most widely utilized. These structure-controlled tFNA nanoparticles undergo a multi-step assembly in response to external stimuli to achieve precise drug release [91]. For instance, four copies of miRNA-29c can be attached to the four vertices of tFNAs via complementary base pairing. In vivo, tFNA enhanced miR-29c delivery and cell membrane penetration, and fully repaired a critical-size skull defect in murine models after 2 months of administration. In vitro experiments showed that stFNA-miR-29c could promote osteogenic differentiation in BMSCs, suppress lipogenesis, and improve cell viability [92]. Also, each vertex of tFNA allows modification of multiple miRNAs at distinct vertices, enabling synergistic regulation of shared biochemical pathways. For instance, S.H. Li et al. modified the 5′ ends of four single-stranded DNAs with sticky ends to obtain s1-4tFNA with four sticky-end apexes, and each sticky end can be hyped with one miR-2661 by DNA/RNA hybridization (Figure 5A). RNase H degrades RNA in RNA/DNA hybrid molecules, only slightly affecting the single- or double-stranded RNA molecules (Figure 5B,E). The stFNA is like a “truck” that transports miRNAs into cells and selects a guide strand to produce subsequent biochemical reactions (Figure 5C). The flow cytometry data showed that the BMSCs had adsorbed a large amount of stFNA during 12 h (Figure 5D) [93]. However, conventionally tFNAs allow miRNAs to bind to either their apical or sidewall, this structure do not adequately protect miRNAs in a complex serum environment. To address this, Xiaoying Lyu et al. developed a tFNA-based bioswitchable miRNA delivery system (BiRDS), where three identical miRNAs are spatially confined to one face of tetrahedron. This design preserves the tFNA’s tetrahedral structure while improving miRNA stability and delivery efficiency (Figure 5F) [94]. tFNAs exhibit intrinsic cell-penetrating capabilities, bypassing the need for any transfection agents to achieve high nucleic acid delivery efficiency [95,96]. Another class of nucleic acid nanomaterials, spherical nucleic acid (SNA)-based vectors, are usually gold nanoparticles in the core and densely loaded with oligonucleotides on the surface. These surface-bound oligonucleotides enable targeted binding and delivery of nucleic acid drugs (Figure 5G) [97].

Despite their promise, nucleic acid vectors face biosafety challenges, including immunogenicity and off-target effects, which require rigorous evaluation for clinical translation. Furthermore, the potential toxicity of long-term in vivo degradation products (e.g., oligonucleotides) requires evaluation, and the cost remains relatively high.

Polyethylenimine (PEI), a cationic polymer that binds electrostatically to negatively charged nucleic acids, is most frequently used in nucleic acids modification. PEI enhances cellular uptake and gene silencing efficiency [98,99]. Although PEI exhibits dose-dependent cytotoxicity, this can be mitigated by grafting negatively charged phenylboronic acid (PBA) groups onto its branched chains or by modifying it with bile acids [100,101]. Branched polyethylenimine (bPEI) variants reduce cytotoxicity while maintaining high transfection efficiency and enabling homogeneous pDNA encapsulation into spherical nanoparticles [102,103]. Meanwhile, the combination of PEI with polyethylene glycol (PEG) reduces its cytotoxicity by decreasing the positive charge of PEI [104,105]. Other cationic polymers, including polybutyl cyanoacrylate (PBCA), poly(β-aminoester) (PBAE), and polylactic acid–glycolic acid copolymer grafted polyethylene (PgP) serve a role similar to that of PEI [106,107,108]. These hydrophilic polymers are often co-used to form core–shell structures that protect nucleic acid stabilization in the core and regulate the release of nucleic acids by adjusting the permeability of the shell [109]. Polylactic acid–glycolic acid copolymer (PLGA) nanoparticles, often prepared via the double-emulsion method, can be surface modified to target specific cells. For instance, modifying its surface with galactose (Gal) effectively enhances PLGA-nanoparticle’s capability to target macrophages [110,111]. PEI-modification of PLGA can enhance its connections with nucleic acid drugs [112].

The high positive charge density of cationic polymers leads to cell membrane disruption, increased lysosomal membrane permeability, called the “proton sponge effect” of PEI, and activation of the mitochondrial apoptotic pathway. Some ligands, such as arginine-modified ones, enhance delivery efficiency but may increase the risk of immunogenicity. To address this limitation, regulating surface charge emerges as a central strategy for attenuating the in vivo toxicity of cationic polymers. By neutralizing the carrier’s positive surface charge, electrostatic interactions with cell membranes are diminished, markedly reducing membrane disruption and cytotoxicity [113].

Polyamide amine (PAMAM) dendrimers, with a highly branched 3D structure, form stable complexes with nucleic acids to enhance intracellular uptake and gene silencing efficiency. Amino-terminated fifth-generation (G5) PAMAMs are commonly used due to their capability to bind pDNA and high affinity for MSCs surface receptor [114,115,116]. To reduce cytotoxicity, PAMAMs are modified with basic amino acids such as histidine, lysine, and arginine. These PAMAM derivatives enhance gene transmission by increasing cell uptake and endoplasmic reticulum escape capacity. These PAMAM derivatives increase endosomal pH via the proton sponge effect, which triggers the endoplasmic reticulum rupture and facilitates the release of pDNA into the cytoplasm [117]. Ganjun Feng et al. have shown hyperbranched polymer (HP), where PEG chain and PEI are attached to the hydrophobic cores, exhibits high pDNA binding affinity and negligible cytotoxicity [118]. Nucleic acid-polymer complexes can also self-assemble into stimuli-responsive micelles, which are capable of changing their structure and performance or achieving controlled release in response to the external environment. For instance, polyplex micelles are formed through the self-assembly of RGD peptide-conjugated with poly(ethylene glycol)-polylysine (PEG-PLys) and pDNA [119]. Alternatively, pDNA can be combined with mixed cationic block copolymers, such as PEG (polyethylene glycol)-b-poly aspartic acid and PNIPAM (poly(N-isopropylacrylamide))-b-poly aspartic acid, at 25 °C. The DET moiety, containing amino groups, facilitates electrostatic interactions with pDNA’s phosphate backbone. These micelles undergo structural reorganization, forming a heterogeneous surface layer composed of hydrophobic and hydrophilic microdomains when heated to 37 °C. This composes thermosensitive polycomposite micelles, which enhance nuclease resistance and protest against nuclease degradation in vivo. Additionally, the optimized surface charge distribution of the micelles reduces non-specific interactions with proteins, further enhancing their biocompatibility and functional efficacy [120].

Notably, inorganic polymers can be designed in both nanoparticles and nanomicelles and further modified through conjugation with diverse molecules. Inorganic nanoparticles used for gene delivery are usually biocompatible. However, their toxic properties can change depending on size and surface coating. It was shown that poor biodegradability and the lack of an excretion mechanism in high-molecular-weight PEI leads to significant cytotoxicity in vitro. A few years after the occurrence, the hypothesis of necrotic cell death during PEI polyplex internalization was proven. The results of the cytotoxicity mechanism study showed that cell membrane destabilization occurs with some cationic polyplex interactions, such as PEI and PLL, in a dose-dependent manner [121]. Currently, optimization of polymer structure, regulation of surface charge, targeted delivery, and development of degradable materials represent core strategies to reduce the in vivo toxicity of cationic polymers. For polymer structure optimization, the design of degradable backbones can be employed: polymers containing cleavable bonds (e.g., ester bonds), such as poly(β-amino ester) (PBAE) and poly(amino co-ester) (PACE), can degrade in vivo and reduce long-term toxicity. PACE polymers reduce systemic toxicity through ester bond hydrolysis while maintaining efficient transfection. Surface charge regulation can reduce electrostatic interactions with cell membranes by neutralizing the positive charge on the carrier surface, thereby significantly mitigating membrane damage and lowering cytotoxicity. Targeted delivery enhances the specificity of carriers for diseased tissues via conjugation with targeting ligands (e.g., RGD peptides, folic acid, and antibodies), which reduces non-targeted accumulation and systemic exposure, consequently decreasing the effective dose and toxicity. The development of novel degradable materials, such as polyether dendrimers and phosphorous-containing dendrimers, minimizes long-term toxicity through their degradable backbones while retaining efficient nucleic acid loading capacity [122]. Their huge potentials in biomedical application warrants further investigation to fully exploit their capabilities.

Mesoporous silica nanoparticles possess hundreds of empty, tunable mesopores capable of encapsulating diverse bioactive molecules, including nucleic acids. The size, volume and structure of these pores can be adjusted according to the size and properties of the cargo dimensions and properties [123,124]. MSNs exhibit low cytotoxicity due to their gradual degradation into orthosilicic acid (Si(OH)₄), which is easily excreted via the kidneys. These properties make them promising candidates for drug delivery. For instance, Radu et al. demonstrated that G2-PAMAM-capped MSNs protect pDNA from enzymatic cleavage while enabling gene expression in HeLa cells [125]. MSN–nucleic acid complexes often exhibit anti-inflammatory properties, and even empty MSNs have also shown anti-inflammatory potential by inhibiting the activation of the NF-κB signaling pathway via TLR2 downregulation. MSNs loaded miR-21-5p promotes angiogenesis and mature vessel formation by targeting SPRY1 and subsequently activating VEGF-induced ERK-MAPK signaling [126]. Yan Li et al. engineered MSNs with amino and trimethylamines to positively charge their surfaces, which stabilizes MSN interaction with negatively charged miRNAs [123]. Stimuli-responsive MSNs, such as disulfide bonded MSN-S-S-NH2, allow MSNs to release loaded miR222 intracellularly by a redox reaction with endogenous GSH [127]. Beyond MSNs, other mesoporous carriers, including magnesium silicate nanospheres with PEI-enhanced oligonucleotide loading capacity [128] and dual-pore structure calcium–silicon nanospheres (DPNPs), are biocompatible, morphology controllable, and bioactive. They can co-encapsulate miRNA-210 and simvastatin (Siv) and significantly accelerate the bone repair process in vitro and in vivo [129].

Gold nanoparticles (GNPs) are another versatile platform for nucleic acid delivery. Biosynthesized using Lignosus rhinocerotis extracts, GNPs exhibit antimicrobial properties [130]. Nucleic acid drugs usually bind to GNPs modified with positive and negative cationic polymers, like PEI, polylysine and PEG-SH (polyethylene glycol-SH). These modifications improve GNPs’ stability, reduce the immune response and minimize non-specific protein interaction [131,132,133]. Similarly, cerium oxide nanoparticles (CNPs) have been widely used in the treatment of oxidative stress-related diseases [134]. Because of oxygen defects in their lattice structures, CNPs can also act as a regenerative free radical scavenger by the redox cycle between the Ce3+ and Ce4+ oxidation states [135], which reduce all harmful intracellular reactive oxygen species (ROS), mimicking antioxidant enzymes like superoxide dismutase and catalase [136]. When conjugated with miRNAs, CNP–miRNA complexes synergistically target both oxidative stress and inflammation [137]. Also, functionalized graphene oxide (GO), surface modified with PEG and PE, enabled efficient miRNA encapsulation via electrostatic interaction and π–π interactions, greatly improving their biocompatibility and anti-inflammatory properties [138].

While inorganic nanomaterials offer immense potential due to their structural and functional plasticity, as well as their tunable physicochemical properties, including size, surface charge, and stimuli responsiveness, their long-term biosafety remains a critical consideration. The long-term bioaccumulation of silica-based nanoparticles, MSNs and CNPs poses significant risks to chronic tissue remodeling, primarily through persistent inflammation, fibrotic progression, and disruption of redox homeostasis. It is reported that accumulation of non-degradable silica in the liver and spleen triggers chronic macrophage activation, which in turn leads to excessive collagen deposition and fibrotic scarring. Long-term studies show that persistent silica residues induce granulomatous lesions, disrupting normal tissue architecture [139]. While CNPs exhibit antioxidant activity through Ce³⁺/Ce⁴⁺ redox cycling, their long-term accumulation can paradoxically induce pro-oxidant effects in specific microenvironments, especially acidic lysosomes. This disruption of cellular redox signaling causes damage to lipids and DNA, and activates senescence pathways. Additionally, nanoparticle aggregates can obstruct renal tubules, reducing filtration capacity and promoting tubular atrophy [140]. Mitigation strategies for these risks may involve the design of rapidly biodegradable MSNs that transform into renally clearable silicic acid, thereby minimizing residual accumulation. For CNPs, surface modification with dextran or PEG has been shown to enhance renal clearance, with 99% of CNPs cleared within 72 h in murine models. Moreover, optimizing CNPs to sizes below 5 nm improves renal excretion and reduces hepatic and splenic sequestration. Similarly, MSNs with enlarged mesopores (>10 nm) facilitate more rapid biodegradation [141]. Bioaccumulation of inorganic nanomaterials poses multifaceted risks to tissue remodeling, necessitating material innovation, such as biodegradable scaffolds, and dosing precision. Future work must prioritize longitudinal toxicity studies and clinically relevant clearance strategies to bridge translational gaps.

Hydrogels enable sustained and sequential nucleic acid release through mechanisms such as degradation, swelling, and diffusion. As the hydrogel degrades, the mesh size expands, allowing nucleic acids encapsulated to diffuse out. Swelling and diffusion. Swelling-induced structure changes in the hydrogel further modulate the release kinetics. Tailoring hydrogel properties, including concentration, physical architecture, the cross-linking method, and cross-linking density, allows precise control over nucleic acids sustained and sequential release [137,152,178]. For instance, Laponite nanoparticles are a kind of synthetic Hectorite with a disc-like structure approximately 25 nanometers in diameter and 1 nanometer thick and charge heterogeneity, which negatively charged surfaces and positively charged edges. These nanoparticles form reversible microporous gel networks, which permit slowly releasing bioactive molecules like miR-22 via face-edge interactions [170]. Similarly, Wang et al. synthesized photopolymerized hydrogels by combining PEG and PLA chains modified by methacrylate (DM) groups. This PEG-PLA-DM hydrogel network gradually reduces cross-linking density by cleavage of PLA ester bonds, loosens the network structure, and accelerates nucleic acid release over time [157]. Release rates are also influenced by nucleic acid-scaffold binding modes, such as electrostatic adsorption, covalent conjugation, and hydrophobic interaction.

Despite these advances, few novel designs are available for sustained release systems. Further research is necessary to optimizer the physical and chemical properties of hydrogels, such as swelling dynamics and degradation rates. More importantly, tissue- and cell-specific release kinetics need to be refined, like how varying release rates impact therapeutic efficacy and potential side effects across diverse cellular environments.

Localized rejecting systems enable site-specific nucleic acid administration, enhancing tissue targeting while minimizing off-target effects. These systems exhibit strong bio-adhesion, allowing conformal coverage of irregular wound surfaces. The advantages can be further enhanced by in situ formed hydrogels, including temperature-sensitive hydrogels and light-curing hydrogels, further optimizing these advantages by leveraging environmental stimuli to achieve precise spatial–temporal control over gelation and drug release.

Photosensitive hydrogels can be classified into photocleavable and photopolymerizable types. Photocleavable hydrogels employ light of a specific wavelength to trigger nucleic acid release; their scaffold materials often incorporate photosensitive moieties, such as o-nitrobenzyl groups or photodegradable cross-linkers. Photopolymerizable hydrogels, on the other hand, undergo a light-induced polymerization of photosensitive components to form a 3D network structure. Photolyzed hydrogel is a highly controllable platform for nucleic acid delivery, leveraging specific wavelength of light to nucleic acid release. These systems are often fabricated using light-responsive hydrogel scaffolds synthesized via Michael addition reactions. Scaffold materials often contain photosensitive groups, such as o-nitrophenyl structure or photodegradable cross-linkers. For instance, Minfeng Gan et al. prepared a photosensitive moiety containing the o-nitrophenyl group to realize the controlled release of miR-26a on demand (Figure 6A). The 3′ end of miRNA-26a was conjugated with the photolytic group (PL-5) containing an o-nitrophenyl structure, enabling ultraviolet (UV)-triggered cleavage and miRNA release (Figure 6B) [148]. Poly(ethylene glycol)-diacrylate (PEG-DA) and PEG-diphotodegradable-acrylate (PEG-DPA) macromers are common hydrogels cross-linked via thiolene click reactions with 8-arm PEG mercaptan (PEG(-SH)8) to form photolabile thiol-acrylate networks (Figure 6C) [181,182,183]. Upon absorption of UV light, the chemical bond undergoes Norrish type I or type II cleavage, and the ester group breaks down into carboxylic acid and alcohol which in turn promotes the degradation of the hydrogel [184]. Notably, the release of nucleic acid is not exclusively light-sensitive hydrogels. siRNAs exhibit baseline diffusion without illumination, while UV accelerates gel degradation [182]. Alternative strategies utilize chemical compound, such as biaryltetrazole (Tet) and or spiropyran (SP) derivatives [185,186]. Tet (II) modified with RGD peptides form hydrogels in PBS at neutral pH, undergoing rapid intra-molecular photo-click reactions under UV light (302 nm) to convert Tet (II) into Tet (II)–GRGD, Tet (II)–GGRGD and Tet (II)–GGGRGD (Figure 6E). Then, Tet (II)–GRGDS transform completely into Pyr(II)–GRGDS within 10 min [186]. Similarly, the photoisomerized spiropyran-galactose (SP-Gal) was synthesized and assembled by linking SP to Gal via a strong π–π stacking of merocyanine (MC) groups upon heating. Visible light irradiation triggers MC-to-SP isomerization, disrupting the hydrogel network, transform the hydrogel into solution and releasing miRNAs (Figure 6D) [185].

While photolyzed hydrogels show promise, most studies remain confined to in vitro models. Critical gaps persist in understanding photodegradation kinetics, biocompatibility, and therapeutic efficacy in vivo. Optimizing light penetration depth, minimizing UV-induced tissue damage, and tailoring wavelength-specific systems for clinical translation are key priorities.

Another type of light-responsive hydrogel is photopolymerized hydrogel. Over the past few years, as a kind of photo-cross-linked biological hydrogel, gelatin methacryloyl (GelMA) hydrogel has emerged as a leading photo-cross-linked biomaterial. Synthesized by adding methacrylate groups to amine residues in gelatin, GelMA incorporates photosensitive groups enabling precise photopolymerization. When exposed to light, the photoinitiator decomposes and releases free radicals, which initiate methacrylamide cross-linking network and eventually form a tunable three-dimensional GelMA (Figure 7A) [188]. This light-responsive gelatin allows precisely spatially controlled nucleic acid delivery to the complex defect site, such as skin wounds and spinal cord injury (Figure 7B) [128,152,155]. The photopolymerization strategy can be applied extensively to chitosan via a nucleophilic substitution reaction. The hydroxyl groups on chitosan react with compounds containing double bonds, such as glycerol methacrylate and GelMA, introducing formyl groups [156]. Riboflavin, a kind of biocompatible photoinitiator, generates radicals when exposed to blue light to cross-link methacrylated polymers like methacrylated glycol chitosan (MeGC) [156,189]. This methodology is adaptable to some other synthetic polymers such as PEG-DA [134,157].

Despite their utility, photopolymerized hydrogels face three key limitations. Firstly, cytotoxicity risks, photoinitiators, like riboflavin, generate free radicals that may leave cytotoxic residues if incompletely consumed during polymerization. More research needs to be applied in developing enzyme-initiated polymerization or metal-free photoinitiators. Secondly, they have limited chemical diversity. Current systems rely mainly on methacrylate groups, with underutilized alternatives like acrylamide or thiol-ene chemistries restricting functional optimization. Exploring norbornene–acrylate hybrids and tetrazole-alkene click chemistry maybe a solution for the development of the next-generation photosensitive system. Thirdly, light penetration remains the principal bottleneck to clinical translation. Clinically, the near-infrared range, UV and blue light, remains limited tissue penetration depths (<2 mm), restricting current photo-responsive hydrogels to superficial sites like skin and cornea or lesions reachable by minimally invasive tools as endoscopically accessible tumors. Raising power to compensate risks collateral photodamage without adequately treating deeper tissue.

To overcome this, four future strategies are emerging: (1) technical synergy—pairing hydrogels with optical fibers, micro-LED implants, or catheter-based endoscopes that deliver light directly to the lesion; (2) alternative stimuli—using more penetrative triggers such as focused ultrasound, magnetic hyperthermia, or low-dose X-ray to indirectly activate photosensitizers or photothermal converters, thereby achieving “remote” light generation in situ; (3) material innovations—engineering ultra-sensitive photosensitizers activatable at very low light fluence and smart carriers that persist at the lesion site and respond to weak local optical signals; and (4) targeted accumulation—enhancing hydrogel density within deep lesions through active targeting (e.g., antibody, peptide, or metabolic ligands) so that therapeutic responses occur under strictly confined, low-intensity illumination.

PH-responsive hydrogels typically exploit dynamic Schiff-base bonds formed between amines and active carbonyl groups. These bonds remain relatively stable under neutral or alkaline conditions but rapidly degrade under acidic environments, enabling targeted delivery and controlled drug release at specific pH values. This property allows targeting delivery via Schiff-base bond breaking at a specific pH value, thus achieving a controlled drug release [192]. For instance, in pathological tissues like tumor where pH value is approximately 5.0, Schiff-base bond-based hydrogels enable targeted drug release (Figure 6F) [187]. In tissue engineering, Yan Li et al. developed an injectable hydrogel (Gel@MSN/miR-21-5p) by cross-linking amino-modified MSN (MSN-NH2-TMA) with aldehyde-modified polyethylene glycol (PEGCHO) and cyclodextrin (α-CD). The Schiff base ensures stability at physiological pH 7.4 but gradually degrades in an acidic environment (pH 6.8), enabling localized miRNA-21-5p release to promote angiogenesis in myocardial infarction [126]. In conclusion, pH-responsive hydrogels are superior for tissue targeting.

Enzyme-responsive hydrogels exploit proteases that are upregulated in injured tissues. For instance, matrix metalloproteinases (MMPs) are highly expressed in osteoarthritis and intervertebral disc degeneration. By incorporating MMP-cleavable peptide sequences into the hydrogel network, the scaffold selectively degrades in response to elevated MMP activity, enabling a stage-wise delivery system. A research shows that the peptide linkage GPLGVRG can be cleaved specifically by a series of MMPs [193]. Building on this approach, the MMP-cleavable peptide CGPLGVRGC was integrated with eight-arm PEG–maleimide solution to enable sustained release of miRNA-29 from MMP-responsive copolymers carriers (Figure 6G). This system achieves a two-stage release: In the first stage, elevated MMP levels trigger the degradation of the hydrogel and release polycomposite micelles. In the second stage, polycomposite micelles undergo MMP-triggered breakdown and facilitate intracellular miRNA uptake and their endoplasmic reticulum escape. When combined into a hybrid hydrogel (miR-29a/PGPC@HG), this strategy significantly attenuated fibrosis in intervertebral disc (IVD) degeneration by targeting miR-29a to diseased tissue [150]. Similarly, hyaluronidase-responsive systems exploit enzyme cleavage of β-N-acetylhexosamine-1,4 glycosidic bonds in hyaluronic acid (HA). An injectable hybrid hydrogel synthesized via the Michael addition reaction of hyperbranched poly(ethylene glycol) diacrylate (HB-PEGDA) and thiolated HA (SH-HA) degrades in osteoarthritic joints, where hyaluronidase and H₂O₂ synergistically trigger sustained drug release from HA-based microparticles [151]. Emerging designs also incorporate dynamic hydrazine bonds and protease-sensitive peptide cross-linkers [145]. Enzyme-cleavable hydrogels rely on the assumption that the target protease is both present and active at the intended site. Their performance, especially those triggered by matrix metalloproteinases, hinges on the actual abundance, activity, and spatial distribution of the target enzyme within the lesion. However, three layers of heterogeneity can compromise this assumption. First, inter-patient variability, like genetic polymorphisms, comorbidities, diet, and microbiome differences, may create patient to patient variation in protease baseline and enzyme expression. Second, enzyme levels rise and fall across disease stages, which may result in transiently overshoot in acute flares, whereas downregulation or compensatory inhibition in chronic lesions. Third, intra-lesion heterogeneity, like spatial gradients in tumor core and invasive rim, produces micro-zones where enzyme abundance differs. These heterogeneities of proteins and enzymes in vivo may result in off-target activation in adjacent healthy tissues with aberrant enzyme expression or negative release when enzyme concentration falls below the activation threshold.

To overcome these challenges, next-generation enzyme-responsive hydrogels can be engineered through three complementary strategies: (1) incorporating cleavage sites for multiple disease-associated enzymes to heighten specificity and mitigate single-marker failures; (2) tuning the activation thresholds of these cutting sites to accommodate the wide range of enzyme activities observed across patients and disease stages; and (3) integrating additional microenvironmental triggers, such as acidic pH, elevated ROS, or altered redox states, to ensure robust activation even when enzyme levels fluctuate. Together, these design principles promise safer and more reliable hydrogels for clinical translation.

Temperature-sensitive hydrogels undergo reversible sol- and gel-phase transition at physiological temperatures, which enable minimal invasive injection and enhance the localized adhesion. For instance, PEG-PLGA-PNIPAM increases the viscosity to form a hydrogel at temperatures above 33 °C and returns to a low-viscosity solution state at room temperature (25 °C). ASP and miR222/MSN can be encapsulated in PEG-PLGA-PNIPAM without structural compromise and retain their spherical structure [127]. The pluronic F-127 (PF-127) hydrogel, an FDA (Food Drug Administration) cleared thermosensitive polymer for clinical use, is a triblock copolymer containing one hydrophobic polypropylene oxide (PPO) chain and two hydrophilic polyethylene oxide (PEO) chains. At lower temperatures, PF-127 molecule tends to form micelles in which hydrophobic PPO chains gather inside the micelles and hydrophilic PEO chains extend outward. At 37 °C, these micelles re-arrange into large-scale organized structures, forming a hydrogel encapsulating shRNA-lentiviral complexes and releasing them persistently (Figure 7C) [20,164,190]. Another triblock copolymer, PLGA–PEG–PLGA, is micelles at low temperature and re-arranges into woven networks at 37 °C via hydrophobic interactions, enabling controlled nucleic acid release (Figure 7D). The nanomicelles formed by this copolymer in water change from a random distribution to a regular linear arrangement at an elevated temperature (37 °C), forming an interwoven network of micelles, which ultimately leads to the curing of the hydrogel. The phase transition caused by this temperature change is due to the fact that the PEG chains are more stretched at lower temperatures and more coiled due to hydration at higher temperatures, which promotes the interaction between the polymer chains and forms a stable hydrogel network, thus realizing the local delivery and controlled release of nucleic acid drugs [102,163].

The cross-linking can be modulated by natural or synthetic hybrids. Hydrogels prepared by mixing chitosan and β-glycerophosphate for delivery of pDNA remain liquid at room temperature and rapidly transform to gelation at b 37 °C by ionic cross-linking between chitosan and β-GP, as well as chitosan chain entanglement, which protecting pDNA structure [81,84]. Similarly, pNIPAAM solution changes from a soluble state to a gelation due to enhanced hydrophobic interactions between pNIPAAM molecules (mainly isopropyl-based interactions) at 32 °C. In this system, siRNAs are loaded on positively charged layered silicates, which forms stable ion exchange with sulfonate groups in pNIPAAM, stabilizing siRNA-loaded hydrogel networks (Figure 7E) [191]. pNIPAAM copolymerized with Dex-PCL-HEMA improves biocompatibility and biofunctionality of the complex [165].

These designs face critical challenges such as strict temperature control because of premature gelation in vivo and thermal instability. Even 1 or 2 °C temperature fluctuation may trigger burst release of nucleic acids and structural collapse.

Unlike stimuli-forming hydrogels, self-assembling hydrogels form stable, ordered structures in situ without external triggers. They leverage intrinsic molecular recognition to spontaneously organize at target sites, maximizing therapeutic precision. These systems comprise two primary classes: the host–guest self-assembly and the peptide self-assembly.

The host–guest hydrogels leverage molecular recognition between β-cyclodextrins (β-CDs) hosts and hydrophobic guests such as adamantane (AD) to form injectable, self-assembling networks. β-CD crystallizes as a pair supported by face-to-face hydrogen bonding and form truncated-cone-shaped cavities stabilized by hydroxyl hydrogen bonding on C2 and C3 (Figure 8A) [194]. The β-CD (host) and AD (guest) self-assemble via hydrophobic interactions to form a gel exhibiting shear-thinning properties during injection, and rapidly self-healing after removal of the shear [144]. Meanwhile, the hydrophobic cavity structure of β-cyclodextrin enable guest encapsulation. Therefore, miRNAs gained hydrophobicity via conjugation with cholesterol or cationic bovine serum albumin (cBSA) enable to bind β-CD cavities to form CD-AD hyaluronic acid (HA). This on-covalent binding also enhances cellular uptake efficiency and miRNA release kinetics (Figure 8B) [144,160,166]. Other hydrogels, such as Pluronic F127, are also suitable for this mechanism [160]. It is worth noting that the self-assembly dynamics and release rates are limited by the ratio and affinity of host–guest molecules, making precise regulation difficult, which requires further in-depth research.

The self-assembled peptides (SAPs) allowing spontaneous formation of the three-dimensional network without chemical or physical treatment makes them versatile tools for minimally invasive surgery and rapid tissue repair. Among the most widely studied self-assembled hydrogel, peptide Ac-(RADA)4-NH2 (RADARADARADARADA-NH2) is a repeating sequence of arginine (R), alanine (A), and aspartic acid (D) arranged as alanine-arginine-aspartate (RAD) units [167,168]. This sequence design facilitates peptide interaction, enabling hydrogel formation with nucleic acids distributed inside (Figure 8C) [167]. In an aqueous environment, RAD peptides favor β-sheet structures due to hydrogen bonding between aspartic acid side chains. The β-sheets structure is critical for self-assembly and hydrogel network formation. Additionally, arginine residues carry positive charges, while aspartic acid residues are negatively charged, further stabilizing the β-sheet structure through electrostatic interactions and promote peptide chains aggregation. Meanwhile, alanine residue provide hydrophobicity, enhancing hydrophobic interaction between peptide chain to reinforce the self-assembled structure [195,196,197]. When exposed to ions, RAD peptide hydrogels undergo ion-induced cross-linking, enabling rapid in vivo gelation while maintaining biocompatibility and degradability (Figure 8D) [168]. In nucleic acid drug delivery applications, SAPs bind nucleic acid drugs via hydrogen bonding and electrostatic interactions, achieving high drug loading capacity (Figure 8D). By non-covalently conjugating the stem cell homing peptide SKPPGTSS (Figure 8C), SAPs can also recruit synovium-derived MSCs (SMSCs) to osteoarthritis injury sites, promoting cartilage regeneration [167,168]. In situ, self-assembled peptide hydrogels mimic the ECM properties, providing a conducive environment for cell attachment, proliferation, and differentiation [168]. This supports localized drug delivery and sustained release. The functionality of these hydrogels depends on precise amino acid sequence design to prevent enzymatic degradation and physical washing in physiological condition.

Based on material composition, morphology and the preparation method, tissue engineering scaffolds can be categorized into three types: non-printed scaffolds for padding, 3D-printed porous braided scaffolds, and 3D scaffolds with special structures.

Scaffolds for padding are designed to address the challenge of irregular tissue damages, particularly bone tissue. Injectable hydrogels, while capable of conforming to irregular cavities, often lack of sufficient mechanical strength. Non-3D-printed scaffolds for padding overcome this limitation by directly filling the defect site. Theses scaffold are typically fabricated from materials like collagen, with lyophilization [209,218]. For instance, S.Elangovan et al. utilized collagen scaffolds to deliver PEI–pDNA complexes to BMSCs, effectively promoting bone tissue formation [206]. However, the inherent mechanical properties of collagen may be insufficient for supporting the repair of large or load-bearing bones. Moreover, the degradation rate of collagen might not align with the pace of bone regeneration, potentially compromising the structural support for newly formed bone. To address these issues, collagen is often chemically cross-linked with other materials, like hydroxyapatite (HA), using lyophilization technology. HA, a highly rigid mineral, possesses the ability to promote osteoblast proliferation and differentiation. When combined with collagen, it significantly improves the mechanical strength and stability of the scaffolds. Additionally, chemical cross-linking improves the scaffold’s pore structure, increases the interconnectivity, and promotes cell adhesion, growth, and differentiation [72,205,207,219]. In summary, through the use of lyophilization technology and chemically cross-linking, non-3D-printed fillable scaffolds can be endowed with good mechanical properties, suitable degradation rate, and osteogenic capability.

3D-printed porous braided scaffolds provide a unique platform for nucleic acids. J. Liu et al. used Direct Ink Writing (DIW) to fabricate β-tricalcium phosphate (b-TCP) scaffolds with tailored pore structures. These scaffolds were subsequently immersed in a solution containing drug carrying nanoparticles (DPNPs), enabling nucleic acid drugs loading (Figure 9A) [129]. Another approach involves encapsulating nucleic acids within hydrogels prior to scaffold loading. H. Hu et al. employed Digital Light Processing (DLP) printing technology and a sintering process to fabricate bioglass scaffolds. They also prepared GelMA/nanoclay hydrogels, soaked the scaffolds in a solution containing miRNA-sEVs, and then applied 365 nm ultraviolet radiation for 2 min to form a coating of sEVs films [216]. Compared with conventional fabrication techniques, 3D printing enables the creation of more complex and precise scaffold architectures. However, the coating method used for nucleic acids loading necessitates prior scaffold preparation. This type of scaffold is relatively uncommon as small molecule nucleic acids and their carriers are loaded physically, which may not ensure sustained and sequential release.

Bionic 3D scaffold design depends on the mechanical properties of target tissues. Morpho-functionally, these scaffolds are classified as 3D fiber–hydrogel or collagen composites or hydrogel composites.

Fiber–hydrogel/collagen composites effectively repair spinal cord and vascular tissues. The spinal cord’s white and gray matter contain orderly nerve bundles within a complex ECM providing mechanical support and cellular regulation. Spinal cord injury (SCI) repair thus requires replicating nerve bundle alignment and ECM properties. N. Zhang et al. utilized aligned polycaprolactone (PCL) fibers to mimic white matter axon arrangement, directing regenerating axon growth. These fibers were embedded in a collagen hydrogel matrix loaded with glial cell line-derived neurotrophic factor (GDNF) and miRNAs (miR-132/222/431). This structure promoted axon regeneration and neuronal functional recovery (Figure 9B–D), with sustained release of the GDNF over 1 week and miRNAs over 3 months, respectively (Figure 9E) [198]. Polymers like PCL, PCL-DPP-PCL, and PCLEEP are often used for the internal fibers, while hydrogels form the outer layer [200,201]. Electrospinning allows the fabrication of nanofibers with specific arrangements and diameters, facilitating cell attachment, migration, and differentiation. This structure is also suitable for vascular tissue engineering. For instance, F. Zhou et al. used PELCL as the inner layer, prepared via emulsion electrospinning and PCL mixed with gelatin were used as the outer layer, prepared by dual-power electrospinning (Figure 9F). The PELCL inner layer delivered miRNA-126 to regulated the response of vascular endothelial cells (VECs), while the outer layer provided mechanical stability and supported cell attachment and growth [199].

Hydrogel composite structure shares similarities to 3D fiber–hydrogel/collagen composites (Figure 9G). B. Khorsand et al. used a fibrin gel as an inner layer to deliver insulin (INS) and active vitamin D3 (VD3) metabolites to improve bone healing in diabetes models without affecting systemic glycemic parameters. The outer layer consisted of a collagen matrix combined with a PEI-(pBMP-2+pFGF-2) nanocomposite. The pDNA encoding BMP-2 and FGF-2 from non-viral gene delivery worked synergistically to promote bone regeneration. The multi-layered structure and multi-drug delivery scaffold enhanced bone formation and healing in diabetes models [99].

While 3D scaffolds with bionic structure offer superior biomimetic properties, their preparation is often complex process. The effectiveness of composites materials requires further investigation. Additionally, producing sheet polymeric fiber scaffolds may involve more complex processes and higher costs compared to conventional drug delivery systems. Although drug release can be controlled by adjusting the coating, precise control in complex biologically environments remain a challenge. The clearance mechanisms of drugs and materials post-release also need to be carefully considered to avoid long-term side effects.

The 3D structure of tissue engineered scaffolds is a key factor in designing nucleic acid release systems. As mentioned previously, injectable hydrogel-based designs are common approach. 3D scaffolds often consist of multiple materials beyond hydrogels. These scaffolds typically have porous structures that facilitate the loading and release of nucleic acids, allowing them to penetrate the scaffold to interact with cells. 3D architecture makes surface modification of scaffolds easier. In addition to traditional physical encapsulation, nucleic acids and their vectors can be adsorbed on scaffolds in the form of coatings.

The nucleic acid release can be designed to align with the tissue regeneration processes. For instance, a collagen sponge can be mineralized in simulated body fluid (SBF) to form a calcium-deficient hydroxyapatite layer. A complex of pDNA encoding bone morphogenetic protein-2 (BMP-2) and polyethylenimine (PEI) is embedded in the CDHA coating. The PEI-pBMP-2 complex to co-precipitate with the minerals in the SBF to form an inner layer. Subsequently, the pDNA encoding fibroblast growth factor-2 (FGF-2) also complexed with PEI is embedded in the CDHA coating to form an outer layer. This double-layer structure design allows for sequential release of the two growth factors. BMP-2 is responsible for promoting osteogenic differentiation of bone marrow MSCs (BMSCs), while FGF-2 is responsible for enhancing cell proliferation and angiogenesis. The sequential release mimic the natural growth factor release pattern during bone healing [203].

In addition to sequential release, 3D scaffolds can also be designed as stimulus-responsive systems, although it is less common. For instance, PHTB (TA-siRNA) hydrogels were synthesized from polyvinyl alcohol (PVA), human-like collagen (HLC), tannic acid (TA), and borax through reactive liquid molding. Borate and hydrogen bonds are crucial for dynamic cross-linking. TA-siRNA nanogels, formed by a self-assembling of TA and siRNA, are loaded into the hydrogel via hydrogen bonds and borax-mediated borate bonds. In diabetic chronic wounds, high oxidative stress due to abnormal glucose and lipid metabolism lead to elevated level of reactive oxygen species (ROS) secreted by macrophages and neutrophils. Borate bonds in PHTB (TA-siRNA) hydrogels are oxidized by hydrogen peroxide at normal physiological pH and temperature. As ROS level rises, borate bonds break, reducing cross-linking density, disintegrating the network structure, increasing pore size, and expanding pore distribution. This change facilitates the release of TA-siRNA nanogels [213].

In summary, 3D nucleic acid drug scaffolds mainly achieve sustained release through self-degradation, pore size, and porosity. However, research on its stimulus-responsive release remains limited. Scaffold topology, particularly pore size and interconnectivity within 3D matrices, critically determines both cellular infiltration and nucleic acid drug release which are critical for therapeutic efficacy. Larger pore diameters (>100 μm) generally facilitate enhanced stem cell recruitment by promoting cell migration, nutrient diffusion, and vascular ingrowth [220]. Conversely, smaller pores (<50 μm) offer higher surface area to volume ratios, increasing nucleic acid loading capacity but potentially restricting cell penetration. This architecture also governs release kinetics: smaller pores create more tortuous diffusion pathways, prolonging sustained release of nucleic acids (e.g., pDNA and siRNA) and protecting them from premature degradation [221]. However, excessive confinement can hinder cell–scaffold interactions essential for transfection. Optimal pore design must therefore balance these competing demands, ensuring sufficient cell ingress while modulating release profiles to match therapeutic windows. Furthermore, pore geometry influences paracrine signaling gradients, indirectly modulating stem cell homing and activation [222]. While 3D structures are effective for localized delivery, implantation induced trauma and complete degradability are concerns requiring further investigation. Future scaffold optimization should integrate computational modeling of pore networks with empirical studies to precisely tailor topology for specific regenerative applications.

The sheet-like scaffolds are primarily composed of hydrogels or polymer fibers. They can be categorized into single-layer and multi-layer types. Additionally, there are distinct layer-by-layer (LbL) thin sheets. An overview of the designs and applications of sheet-like delivery systems is shown in Table 3.

Monolayer hydrogel sheets serve as critical platforms for skin tissue repair, providing physical infection barriers when applied directly to wounds. For instance, S.C. Tao et al. developed chitosan hydrogel dressings loaded with miR-126-3p-enriched synovial MSC exosomes (SMSC-exos). These enabled sustained exosome release and accelerated healing of full-thickness diabetic wounds in rats. Beyond establishing moist microenvironments, the chitosan matrix contributed essential hemostatic and antimicrobial functions, particularly vital for chronic diabetic wound healing [52]. To address deeper tissue delivery, microneedle patches have been explored for transdermal administration. These patches are able to penetrate the stratum corneum, facilitating deeper drug delivery with minimal pain [232,233]. M. Qu et al. reported a transdermal microneedle patch for delivering PBAE/DNA nanoparticles through a cross-linkable GelMA matrix. The GelMA microneedle patches can penetrate the epidermis to reach the targeted cells (Figure 10A) [107]. The performance of sheet hydrogels can be enhanced by combining with other polymers. For instance, C. Cai et al. created a composite film by integrating hydrogels with PCL (polycaprolactone) electrospun nanofibers to form that enhanced adhesion to peritendinal tissues (Figure 10B) [223]. In another study, scaffolds combing a collagen-chitosan layer with a silicon membrane can control water loss and prevent bacterial entry, winning time for transplantation of ultra-thin skin grafts [225].

In summary, hydrogel sheets are of great significance for skin tissue engineering. However, further research is needed for their effectiveness, particularly in reducing scarring while promoting skin regeneration.

Polymer fiber 3D sheet-like scaffolds are primarily created by electrospinning, which involves melting a polymer to solution and ejecting it with a high voltage to form nonwoven fiber pads. This technology is widely used in tissue engineering and biomedicine [234]. A. Malek-Khatabi et al. fabricated a polycaprolactone (PCL) nanofiber scaffold for the delivery of chitosan–pDNA complexes by electrospinning. The pDNA and chitosan were mixed and modified by a microfluidic mixer to form nanocomplexes. This nanofiber structure provides a large surface area for pDNA loading and supports cell growth and penetration in vivo and in vitro (Figure 10C) [86]. The surface of these scaffolds can also be easily modified by polymers and hydrogels. For instance, Y. Wang et al. prepared PLLA (poly-L-lactic acid) and silk fibroin parallel fiber (PSPF) films via electrospinning and modified their surface with PGA. The PGA coating gave PSPF@PGA fiber membrane a negative charge, facilitating its binding to RCP/pDNA nanoparticles through electrostatic attraction. PGA coating may also improve the membrane’s compatibility with cells, promoting cell adhesion and proliferation. This patch-style scaffold serves as a local gene delivery platform for delivering the pJUN plasmid to Schwann cells in damaged nerve regions [228].

Layer-by-layer (LbL) thin sheets are another type of sheet-like scaffolds. This technique involves constructing multi-layer thin sheet structures by alternating layers of oppositely charged materials, allowing for precise control of the thickness and composition. Mandapalli, P. K et al. created an LbL film by alternately adsorbing chitosan and sodium alginate solutions on a glass template, forming a 15-layer bilayer structure loaded with EGF and TGF-β siRNA. The film exhibited excellent mechanical properties, supported the adhesion and growth of A431 epidermal keratinocytes, and inhibited bacterial growth in vitro. In a C57BL/6 mouse incision wound model, the film accelerated wound healing, reduced collagen deposition and significantly promoted re-epithelialization by releasing constantly EGF and TGF-β [231]. In another study, siRNAs were encapsulated on dressings using LbL technology. The scaffolds consisted of a degradable layer poly(β-aminoester)2 (Poly2), then dextran sulfate (DS) in the middle, and an upper siRNA-containing chitosan layer. Poly2 is a well-established hydrolytically degradable polycation that facilitates water-based erosion in multi-layer films. The anionic DS and the cationic chitosan form stable multi-layer membrane structures through electrostatic interactions (Figure 10D–F). Poly2, containing β-amino ester groups with strong cationic properties, forms complexes with negatively charged chondroitin sulfate. Its molecular chain includes ester bonds (-COO-) and amino ester bonds (-NH-CO-O-), which hydrolyze in physiological environments for the slow siRNAs release. This structure effectively protects siRNA and enables precise membrane construction. LbL technology protects siRNAs during delivery, reducing digestion risks [235]. Recently, LbL strategies have also been extended to effective intracellular mRNA delivery, a topic of growing interest. For example, positively charged poly-L-lysine (PLL) and poly-L-ornithine (PLO) form mRNA polyplexes that are arranged in an LbL fashion with negatively charged layers. These assemblies are stable at extracellular pH (remaining net negatively charged), but they acquire a net positive charge in the acidic endosomal environment. This pH-triggered charge reversal facilitates endosomal membrane disruption and promotes cytosolic release of mRNA [236]. LbL thin sheets are advanced scaffolds that require further research to fully explore their potential.

The design of nucleic acid release for sheet-like scaffolds is another important aspect to consider. The swelling and degradation of hydrogels have been discussed earlier in Section 5.1.1, and similar principles apply to sheet-like hydrogels as injectable hydrogels. Sheet hydrogels can sometimes be designed to be stimulus responsive [86,223]. For instance, A. Malek-Khatabi utilized chitosan as a cationic carrier to form nanocomposites (NCs) with plasmid DNA (pDNA) through electrostatic interactions. These NCs were then linked to the MMP-sensitive peptide sequence (GCRDGPQG↓IAGQDRCGC) on the surface of PCL nanofibers using cross-linker Sulfo-LC-SPDP. This design allows for the substantial release of NCs in an environment containing MMP-2 enzyme [86]. The release mechanisms for both hydrogels and polymeric fiber sheet-like scaffolds are primarily based on degradation. In particular, multi-layer sheet-like scaffolds can achieve the continuous and sequential release of nucleic acid drugs through LbL degradation [237]. S. A. Castleberry et al. designed a degradable vector system consisting of a poly(β-aminoester)2 and dextran sulfate (DS) bottom layer, and an upper siRNA-containing chitosan (Chi) layer. In this system, the release of siRNAs is controlled by adjusting the number of layers of the LbL membrane, and the degradation rate of the underlying layer determines the release rate and duration of siRNA [235]. In general, there are relatively few designs for sustained nucleic acid release from sheet-like scaffolds. Future research should focus on developing more advanced sheet-like 3D scaffolds to achieve better control over the nucleic acid release characteristics.