Drug delivery systems: current applications and future directions

Wait 5 sec.

IntroductionDrug delivery systems (DDS) have evolved from formulation aids into determinants of therapeutic feasibility,1 reflecting changes in the types of therapeutic agents. Early drug development was dominated by small molecules with favorable pharmacokinetics and well-defined systemic exposure, for which conventional dosage forms generally sufficed.2 In contrast, the rise of biologics, including monoclonal antibodies, recombinant proteins, and peptides introduced constraints of molecular stability, enzymatic degradation, and limited permeability across biological barriers, such as the gastrointestinal tract and blood-brain barrier (BBB).2,3,4,5 Rather than being a secondary consideration, drug delivery is now a central component of therapeutic design. This transition has been accompanied by successive technological advances, which have progressively expanded the capabilities of therapeutic delivery from conventional formulations to targeted, nucleic acid-based, and living delivery systems. The major technological and clinical milestones that have shaped this evolution are summarized in Fig. 1.Fig. 1Full size imageEvolution of drug delivery systems from rational formulation design to functional therapeutics. Drug delivery systems (DDS) have developed in parallel with advances in therapeutic modalities and changing clinical requirements. The timeline summarizes major stages in this progression, beginning with formulation technologies designed to improve drug solubility, stability, circulation time, and controlled release. Subsequent developments introduced long-circulating nanoparticles, active targeting strategies, including antibody-based delivery, and intracellular delivery platforms, followed by technologies capable of delivering nucleic acids, genome-editing systems, and other complex biological therapeutics. Representative regulatory approvals are shown to illustrate the clinical translation of these technologies. Years shown in the timeline indicate the first FDA approval of each representative product. Early examples, including Doxil® and Abraxane®, demonstrated that modification of drug distribution could improve tolerability while maintaining therapeutic activity. Later approvals, such as Kadcyla®, established targeted delivery using antibody–drug conjugates, whereas Onpattro® and lipid nanoparticle–based mRNA vaccines demonstrated the clinical feasibility of non-viral intracellular delivery. More recent approvals, including Zolgensma®, further expanded the therapeutic scope of DDS by enabling gene replacement therapy. The lower panels summarize how the objectives of delivery design have changed over time. Initial efforts primarily addressed pharmacokinetic limitations through improvements in drug stability, circulation, and controlled release. Subsequent advances focused on specific delivery of therapeutic cargo through tissue targeting, intracellular transport, and controlled activation at target site. These technological developments also broadened the clinical application of DDS beyond oncology to neurological, cardiovascular, metabolic, infectious, genetic, and autoimmune diseases. The timeline illustrates that the role of DDS has gradually shifted from improving the disposition of existing drugs to enabling therapeutic modalities that depend on efficient and selective delivery. Contemporary delivery systems, therefore, function not only as carriers but also as essential components that determine therapeutic activity, tissue specificity, and clinical applicability. Created with BioRender.comDrug delivery has traditionally been classified by the materials used, such as liposomes, polymeric nanoparticles, or viral vectors.6,7,8,9 This composition-based approach describes formulation strategy,1,2 but does not capture in vivo behavior, which depends on the payload, route, and interactions with physiological barriers. Systems of similar materials can behave differently, whereas systems with distinct compositions can converge on the same functional outcomes.7,8,9 These limitations have become more apparent with therapeutic goals shifting toward diseases with complex pathophysiology, including cancer, neurodegenerative disorders, autoimmune diseases, and genetic conditions, in which systemic exposure alone is insufficient. 10,11,12,13Recent reviews have examined individual delivery platforms in depth but remain organized around specific materials or applications;14 therefore, cross-platform comparisons are limited and generalizable design principles are inconsistently defined.15,16 A framework spanning modalities and linking engineering parameters to biological constraints remains underdeveloped.17,18 Further, omission of direct translation cost needs to be addressed, because it obstructs the rational transfer of design principles between systems and their selection in the clinic.11,19Over the past two decades, the therapeutic landscape has expanded to include nucleic acid drugs such as small interfering RNA (siRNA), messenger RNA (mRNA), antisense oligonucleotides, and CRISPR-associated constructs,20,21,22,23 cell-based therapies, engineered bacteria, and antibody-drug conjugates (ADCs),11,24,25,26 with lipid nanoparticles (LNPs) considered the leading non-viral platform for nucleic acid delivery.27,28 These modalities and their distinct delivery requirements are examined in the subsequent sections.These advances cover therapeutic areas in which delivery constraints directly influence treatment outcomes. In solid tumors, heterogeneous vascularization and elevated interstitial pressure limit drug penetration.29 In central nervous system disorders, the BBB restricts systemic access.30,31 Immune-mediated diseases necessitate spatial control of immune modulation.32,33 These disease contexts, and the delivery strategies developed to address these, are examined in detail in the disease-oriented clinical section.Despite these advances, delivery systems continue to be grouped by material.1 LNPs, ADC, viral vectors, and cell-based carriers are often discussed as independent categories despite addressing overlapping challenges,34,35 including circulation stability, tissue targeting, cellular uptake, and intracellular trafficking.36 Therefore, a material-based scheme precludes direct comparison and obscures shared design principles, a gap underscored by how little of the administered material typically reaches its target.37To transcend this view, this review classifies delivery systems along three independent dimensions. The first is carrier identity, separating synthetic systems from living platforms. Synthetic systems, including LNPs, polymeric carriers, and inorganic materials, provide control over physicochemical properties and manufacturing.38 Living carriers, including cells and engineered microorganisms, utilize endogenous biological functions such as migration, immune interaction, and environmental sensing.24,25,26,39 The second dimension is payload class, small molecules, biologics, nucleic acids, and cell-based therapies,40 each with distinct requirements regarding stability, delivery route, and intracellular processing.23,28 The third is the targeting mechanism, which spans passive, active, and programmable strategies.16,41 The delivery systems included in this review were selected based on their clinical relevance and their representation of mechanistically distinct approaches to drug delivery.This three-dimensional framework provides a consistent basis for comparison. By separating composition from function, it allows evaluation of systems based on their biological objectives and exposes challenges common to different platforms, which in hindered by a material-based grouping.16 Viewed through this lens, platforms apparently unrelated by composition prove to share the same rate-limiting step—LNPs, viral vectors, and bacterial carriers must each overcome intracellular delivery barriers to release their payload, just as ADCs and engineered cells both succeed or fail regarding the target recognition specificity. For instance, LNPs and viral vectors must both achieve endosomal escape by different mechanisms,28,42,43 whereas ADCs and cell-based systems both depend on selective targeting through distinct biological processes.11,24,44This review focuses on delivery systems whose architecture is the primary functional element in the localization, transport, or controlled release of therapeutics in vivo. These include lipid-based systems, polymeric and inorganic nanoparticles,45 targeted molecular conjugates,46 viral and non-viral vectors,47 and cell- and bacteria-mediated platforms.48 Complementary approaches, including extracellular vesicles, protein-based carriers, hydrogels, device-assisted delivery systems, and prodrug strategies, have been included for completeness,49,50,51 whereas agents whose effect is derived from intrinsic pharmacology rather than delivery architecture are beyond its scope.On this common basis, advances in nanomedicine,11 biomolecular engineering,52 and synthetic biology53 can be evaluated together rather than in isolation. The review is organized to follow this framework. We first trace the clinical evolution of DDS and their expanding role across diseases, then describe the major modalities within the abovementioned taxonomy, and finally provide a mechanistic account of the multiscale biological barriers shaping delivery efficiency. We then examine clinical applications from a disease-oriented perspective, and close with emerging directions in computational design and integration, followed by a unified synthesis of shared translational bottlenecks—spanning biological complexity, safety, manufacturing and reproducibility, and clinical deployment.Clinical evolution of drug delivery systems and regulatory milestonesDDS have developed alongside advances in therapeutic agents, clinical requirements, and regulatory expectations. Early studies on drug delivery focused on improving the pharmacokinetic properties of small molecules, including solubility, stability, and systemic exposure. Several small-molecule drugs showed activity in vitro but failed to achieve adequate distribution or persistence in vivo.54 Lipid-based carriers, particularly liposomes, were among the first systems to be used clinically, enabling encapsulation of hydrophobic compounds and altered biodistribution, resulting in reduced off-target toxicity, especially in oncology.55 Doxil, the first FDA-approved nanomedicine (1995), exemplified this advance.56 Polymeric carriers subsequently offered greater control over size and release,57 whereas PEGylation reduced protein adsorption and extended circulation;58 however, its long-term immunogenicity is now itself under scrutiny.59 At this stage, delivery systems primarily supported existing drugs rather than enabling new therapeutic classes.Most early clinical applications were concentrated in oncology.60 Cytotoxic agents have narrow therapeutic windows, and delivery systems provided a means to alter biodistribution and reduce exposure to non-target tissues. Passive targeting via the enhanced permeability and retention (EPR) effect increased accumulation in tumor tissue;61 despite the inconsistency of the EPR effect across tumor types and patient populations,62 it served as a practical design principle for early nanomedicine, facilitating the development and early clinical acceptance of liposomal and polymer-based formulations.As therapeutic agents grew more complex, delivery requirements shifted accordingly. Conjugation strategies such as PEGylation and fusion proteins were developed to stabilize protein and peptide therapeutics and extend their circulation.63 These approaches were not confined to oncology: delivery systems were increasingly applied in inflammatory, metabolic, and endocrine disorders for extending dosing intervals, maintaining stable exposure, and reducing immunogenicity.64,65Nucleic acid therapeutics posed a different delivery problem: unlike earlier agents, siRNA, mRNA, and antisense oligonucleotides needed to survive in circulation as well as reach the cytosol to act, a requirement that conventional carriers could not meet.66 The clinical arrival of this class was therefore inseparable from advances in delivery. The approval of Onpattro (patisiran), the first siRNA, LNP therapeutic in 2018, and the subsequent authorization of mRNA vaccines established that non-viral systems could deliver nucleic acids in vivo at scale; the growing roster of approved oligonucleotide drugs has since consolidated this shift.67 Thus, delivery systems moved from modifying pharmacokinetics to enabling entirely new therapeutic mechanisms.Viral vectors have played a distinct role in the evolution of DDS. Adeno-associated viruses (AAVs) enable long-term expression of therapeutic proteins and now underpin several approved gene therapies.68,69 Their principal constraints, including host immune responses, vector persistence, and manufacturing complexity, are fundamentally different from those of synthetic carriers,70 and they show differences in tissue tropism and dosing shape vector selection. Consequently, viral approaches continue to advance in parallel with non-viral approaches rather than as their substitutes.Targeting strategies evolved from passive accumulation toward ligand–receptor recognition for greater specificity. ADCs exemplify this shift, as they pair antigen-directed targeting with cytotoxic payloads to achieve selective delivery in multiple cancers.71,72 Peptide–drug and small-molecule conjugates have since extended targeted delivery to additional targets with improved tissue penetration and simpler synthesis,73,74 resulting in improved localization while limiting exposure to non-target tissues.Nonetheless, the clinical performance of active targeting has been variable. High binding affinity at the cellular level does not reliably translate to greater tissue-level accumulation, a mismatch first formalized as the “binding-site barrier”;75 vascular structure, interstitial pressure, receptor accessibility, and intratumoral heterogeneity further limit ligand-directed delivery.76 Consequently, recent designs pair targeting with additional mechanisms such as controlled release and environmental responsiveness, using systems that react to pH, enzymatic activity, or other local cues for improved context-specific delivery.77Delivery systems are now applied well beyond oncology, an expansion that is mirrored in regulatory approvals spanning genetic, infectious, cardiovascular, metabolic, and immune-mediated diseases. Each context has distinct constraints: crossing the BBB in CNS disorders,30 limiting systemic exposure in cardiovascular and metabolic disease,78 and achieving spatially localized immune modulation in immune-mediated conditions.79 These disease-specific requirements, detailed in the clinical applications section, underscore that delivery strategy must be matched with the biology of each target setting.Combining delivery approaches into hybrid systems that sequentially address multiple barriers has also attracted interest. For example, in cell membrane–coating, synthetic nanoparticles are cloaked in natural cell membranes, conferring immune evasion and prolonged circulation while retaining the tunability of the synthetic core,80 an approach since extended across cell types for targeted delivery.81 In modular designs, distinct components independently handle circulation, targeting, and intracellular release, allowing each function to be optimized separately.82Regulatory approvals mark the clinical trajectory of delivery systems. Early approvals were centered on liposomal and polymer-based formulations in oncology, where altered drug distribution improved the tolerability of cytotoxic agents,56 followed by PEGylated proteins and conjugate systems that extended circulation and broadened biologic use.58 More recently, delivery has become inseparable from therapeutic function as LNPs have enabled the clinical use of siRNA therapeutics and mRNA vaccines,83 viral vectors have gained approval for gene-replacement therapies,69 and antibody–drug conjugates have expanded across multiple cancers.84 Together, these milestones reflect a shift from delivery as a supportive strategy to delivery as an essential determinant of therapeutic activity.With growing complexity, the manufacture and scalability of delivery systems have become decisive constraints. Liposomal and polymer-based systems have relatively established production, whereas nucleic acid carriers and cell-based therapies require more controlled, reproducible processes, as formulation variability directly affects stability, activity, and safety.85 Standardized production and characterization are thus prerequisites for clinical translation,69 and regulatory evaluation increases in complexity as systems incorporate multiple interacting components.86 When viewed across these developments, drug delivery has progressed along several axes, from small molecules to biologics and nucleic acids, from passive to targeted and responsive delivery, and from single-disease to broad clinical use, culminating in a shift from formulation-centered optimization to system-level design, wherein delivery is integrated with therapeutic function and shaped by shared biological and translational constraints.Engineering modalitiesContemporary DDS encompass a broad range of engineering strategies with distinct mechanisms and clinical applications. For clarity and consistency, this section is organized into four categories: synthetic systems, targeted molecular conjugates, biological delivery systems, and complementary platforms (Fig. 2). Each category is discussed using a common framework considering design principles, mechanisms of action, clinical maturity, and current limitations.Fig. 2Full size imageUnified classification framework for contemporary drug delivery systems. Contemporary drug delivery systems (DDS) encompass a wide range of carrier platforms that differ in material composition, biological origin, therapeutic cargo, and mechanism of action. This figure organizes these platforms using six complementary criteria: carrier type, representative platform, payload class, targeting mechanism, functional advantage, and clinical translation. The resulting classification provides a common structure for comparing delivery systems that have traditionally been described separately. The upper panel categorizes delivery systems into four major groups: synthetic carriers, biological carriers, bioconjugates, and physical or device-assisted delivery systems. Representative examples are shown for each category, illustrating the diversity of currently available delivery technologies. Although these platforms differ substantially in composition and engineering strategy, many have been developed to address similar biological and pharmacological challenges. The middle panels summarize the principal payload classes and targeting mechanisms associated with different delivery systems. Therapeutic cargo ranges from conventional small molecules and proteins to nucleic acids, genome-editing systems, and living cells. Targeting strategies include passive accumulation, ligand-mediated recognition, tropism-driven delivery, antigen-specific targeting, and externally or environmentally responsive activation. Individual platforms frequently combine multiple targeting mechanisms according to the biological requirements of specific therapeutic applications. The lower panels highlight functional characteristics that influence delivery performance across different platforms. These include protection of therapeutic cargo, prolonged circulation, improved tissue penetration, intracellular transport, target specificity, controlled release, and reduction of systemic toxicity. Representative clinically approved products illustrate how these design principles have been translated into therapeutic practice. Common engineering considerations, including biocompatibility, physicochemical stability, manufacturability, reproducibility, functional efficacy, immunogenicity, and regulatory feasibility, remain relevant regardless of carrier type and continue to guide the development of next-generation delivery systems. Created with BioRender.comSynthetic drug delivery systemsSynthetic DDS are among the most clinically advanced platforms, rationally engineered from well-defined materials with tunable size, charge, composition, biodegradability, and release kinetics to overcome barriers of circulation, clearance, penetration, and intracellular-delivery. Across synthetic platforms, a shared set of physicochemical parameters governs in vivo fate.87,88 Particle size dictates clearance; systems approximately 10–200 nm in diameter circulate the longest and exploit EPR-mediated tumor accumulation, whereas larger particles are removed by the mononuclear phagocyte system (MPS), and those below approximately 10 nm are eliminated by renal filtration.89,90 Surface charge shapes biological interactions: cationic surfaces enhance uptake but accelerate opsonization, whereas neutral or slightly anionic surfaces are generally favorable for systemic use. Morphology and surface chemistry (e.g., PEGylation) further tune margination and immune evasion.91,92 These determinants recur across the lipid, polymeric, and inorganic systems described below, which are therefore expanded with emphasis on their distinguishing features.Lipid nanoparticles (LNPs)LNPs are the most widely used non-viral platform for nucleic acids, small molecules, peptides, and proteins, and have shown clinical success in approved siRNA drugs and mRNA vaccines.28,93,94,95 Encapsulation protects nucleic acids from degradation and enables efficient intracellular delivery compared with that of viral vectors. LNPs are non-integrating, chemically tunable, and suitable for scalable manufacturing. Clinically relevant formulations comprise ionizable lipids, helper phospholipids, cholesterol, and PEGylated lipids, each contributing distinct physicochemical properties that together determine biodistribution, stability, and endosomal escape efficiency (Table 1).96,97,98,99Table 1 Key features, innovations, and challenges of lipid nanoparticles (LNPs)Full size tableDesign and delivery mechanismIonizable lipids are central to LNP function. Near-neutral at physiological pH, they minimize serum-protein binding and systemic toxicity, whereas they protonate under acidic conditions, forming complexes with nucleic acids and drive intracellular release.93 Their acid dissociation constant (pKa) is a critical design parameter. A pKa of approximately 6.2–6.5 provides optimal in vivo nucleic acid delivery, as exemplified by the clinically approved ionizable lipid DLin-MC3-DMA (pKa ≈ 6.4).100,101 Machine-learning models trained on LNP formulation data have since confirmed the importance of this pKa window and of ionizable-lipid substructure for mRNA-vaccine potency.102 The nitrogen-to-phosphate (N/P) ratio governs electrostatic complexation between lipid amines and nucleic-acid phosphates. A ratio that is too low impairs encapsulation, whereas one that is too high increases cationic charge and cytotoxicity; thus, an N/P ratio of approximately 3–6 balances encapsulation efficiency and biocompatibility.103Helper lipids and cholesterol tune structure and pharmacokinetics. Saturated phospholipids such as DSPC rigidify the bilayer and support fusion during endocytosis, while cholesterol modulates packing and fluidity, thus improving encapsulation and prolonging circulation.104,105,106,107,108,109 Excess cholesterol, however, can impair endosomal release. PEG-lipids form a hydrophilic corona that limits aggregation and opsonization, but repeated dosing can induce anti-PEG antibodies, which has motivated the use of cleavable PEG-lipids and alternative hydrophilic polymers.93After endocytic uptake, LNPs are entrapped in endosomes, and the cargo must escape into the cytosol before lysosomal degradation. This step is strikingly inefficient, with only under 2% of internalized siRNA estimated to reach the cytosol, making endosomal escape the principal barrier to intracellular delivery;99 therefore, its improvement through amino-lipid and endosomolytic design remains a central engineering goal.106,107 Two non-mutually-exclusive mechanisms are proposed. In membrane destabilization, protonated ionizable lipids interact with anionic endosomal phospholipids to form inverted-hexagonal (HII) phases that rupture the membrane; the propensity for this transition correlates with pKa ~6.2–6.5, but lipids of similar pKa can differ markedly in transfection, indicating that headgroup flexibility, linker chemistry, tail structure, and cell-specific trafficking also govern escape.42 In the complementary proton-sponge mechanism, buffering lipids drive the influx of protons, chloride, and water during acidification, resulting in osmotic swelling and endosomal rupture.110,111Endosomal escape is increasingly considered a biophysical event coupled with innate immunity. Lipid-mediated membrane perturbation leaks endosomal contents into the cytosol, which the cell interprets as a stress signal that activates pattern-recognition receptors including the NOD-like receptor pyrin domain-containing protein 3 inflammasome;112 meanwhile, the nucleic-acid cargo is sensed by endosomal toll-like receptor (TLR) 3, TLR7, and TLR8 as well as by cytosolic retinoic acid-inducible gene-I and melanoma differentiation-associated protein 5, driving type-I interferon and pro-inflammatory responses.113 This intrinsic immunostimulation is beneficial in vaccines, providing a built-in adjuvant that enhances antigen presentation, but excessive activation causes reactogenicity. Therefore, LNP design must tune immune activation within a controlled window.83As escape is difficult to quantify and total uptake overestimates functional cytosolic delivery, several orthogonal assays are used. Galectin-recruitment imaging (e.g., Gal8-GFP) reports endosomal membrane damage as punctate signals, but reflects damage rather than productive release.114 Functional reporters like split-luciferase systems (e.g., the SLEEQ platform, in which HiBiT-tagged cargo complements cytosolic LgBiT) provide a direct, high-throughput readout of productive escape,115,116 while subcellular fractionation biochemically separates cytosolic and vesicular cargo but is limited by incomplete organelle separation. As each method carries distinct biases, combining orthogonal approaches is generally necessary for reliable assessment.117Biodistribution and lipid engineeringPronounced hepatic tropism is a major intrinsic limitation: conventional LNPs accumulate in the liver via ApoE-mediated hepatocyte uptake, restricting extrahepatic delivery.108,118 Selective organ targeting (SORT) redirects distribution to the spleen, lung, or immune cells by adding charged lipids. For instance, piperazine-containing ionizable lipids favor antigen-presenting cells for immune-oncology and vaccination, whereas branched tails and heteroatom substitutions shift tropism toward lymphatic tissue or the central nervous system.119,120Degradable ionizable lipids further reduce hepatic retention by accelerating nanoparticle disassembly after uptake, which shortens intracellular residence in hepatocytes and Kupffer cells. Hydrolysable ester, acetal, or carbamate bonds are cleaved by esterases and hepatic lipases, yielding lipid fragments that enter endogenous metabolism. Ionizable lipids are then broken into fatty acids, alcohols, and amine species; helper lipids and cholesterol join physiological lipid pools; and PEG-lipids dissociate and are cleared through the renal or hepatobiliary system.109,121 Rational design that pairs biodegradability with a predictable metabolic fate is thus essential to balance delivery efficiency against long-term biosafety.Clinical translationThe clinical utility of LNPs was first demonstrated with patisiran, an siRNA therapy for hereditary transthyretin amyloidosis, which delivers its cargo to hepatocytes and produces sustained transthyretin reduction, making it the first systemically administered RNAi agent to reach the clinic.122 Global deployment of the COVID-19 mRNA vaccines BNT162b2 and mRNA-1273 then confirmed the platform versatility and large-scale, safe manufacturability,123 thus accelerating the broader clinical translation of nanomedicine.83 Applications now extend to oncology, through mRNA-encoded tumor antigens and oncogene-silencing siRNA; to in vivo CRISPR-Cas9 editing, with early clinical gene editing achieved in transthyretin amyloidosis;124 and increasingly, to extrahepatic and central nervous system targets, where engineered LNPs are being developed to deliver mRNA across the BBB for neurodegenerative indications.125,126,127 A modality-specific limitation is anti-PEG immunogenicity, wherein repeated dosing induces anti-PEG antibodies that accelerate clearance and can cause hypersensitivity.128Polymeric nanoparticles and micellesPolymeric nanocarriers, comprising nanoparticles and micelles, are versatile, clinically validated platforms whose tunable structure, biocompatibility, and capacity for controlled and targeted release have expanded precision-medicine applications in oncology, inflammatory disease, and nucleic acid therapy (Table 2).129,130,131Table 2 Summary of polymeric nanoparticles and micellesFull size tableClassification and structurePolymeric nanocarriers are fabricated from biodegradable synthetic polymers such as poly (lactic-co-glycolic acid) (PLGA), polylactic acid (PLA), polycaprolactone (PCL), PEG, and amphiphilic block copolymers, or from natural polymers such as chitosan, alginate, and hyaluronic acid, and divide into two structural classes.132 Polymeric nanoparticles are solid colloidal systems (10–1000 nm) comprising nanospheres, in which the drug is dispersed throughout the matrix, and nanocapsules feature a polymeric membrane surrounding a drug-containing core. Both release cargo through diffusion, polymer degradation, or surface/bulk erosion, and their rigidity confers high physicochemical stability for delivering chemotherapeutics, proteins, vaccines, and nucleic acids.133,134,135 Polymer selection governs the degradation and release profile: PLGA hydrolyzes into lactic and glycolic acids at rates tunable by the lactide-to-glycolide ratio and molecular weight, enabling release over days to months, while natural polymers such as chitosan, alginate, and hyaluronic acid add intrinsic biocompatibility, mucoadhesion, and receptor-targeting capacity (e.g., hyaluronic acid binding CD44 on tumor cells) but at the cost of greater batch-to-batch variability.136,137,138Polymeric micelles are self-assembled core–shell aggregates formed by amphiphilic block copolymers above the critical micelle concentration.139 A hydrophobic core solubilizes poorly water-soluble drugs while a hydrophilic corona, frequently PEG, reduces opsonization and prolongs circulation; micelles are typically smaller (10–100 nm) with improved solubilization and circulation relative to those of solid nanoparticles, but can dissociate upon dilution below the critical micelle concentration, causing premature drug release.51,140,141,142 As with other synthetic carriers, biodistribution is governed by size, charge, and morphology (see Synthetic intro), but polymeric systems illustrate these determinants concretely where particles of 10–200 nm exploit prolonged circulation and EPR-mediated tumor accumulation while avoiding rapid renal filtration, and non-spherical geometries such as rods and worm-like structures alter margination, cellular internalization pathways, and tumor penetration relative to spheres.91,143,144,145 Surface charge presents a characteristic trade-off. Cationic surfaces enhance electrostatic uptake and are advantageous for delivering large, negatively charged nucleic acids, but also promote serum-protein adsorption, complement activation, and accelerated clearance; therefore, neutral or slightly anionic surfaces are generally preferred for systemic administration. A further limitation is the low drug-loading capacity for hydrophobic drugs such as paclitaxel and camptothecin, often below 5–10% (w/w), which increases the carrier burden and motivates prodrug-based polymers and stimulus-responsive designs.146Clinical validationSeveral polymer systems are clinically established. PLGA-matrix depots, including Lupron Depot (leuprolide), Risperdal Consta (risperidone), and Zoladex (goserelin), provide sustained, controlled release, demonstrating the clinical feasibility of polymer-based delivery.147,148,149 Polymeric micelles have advanced particularly in oncology: Genexol-PM, a mPEG-PDLLA paclitaxel micelle, was the first clinically approved polymeric micelle (South Korea), which eliminated the Cremophor EL solubilizer to reduce infusion-related hypersensitivity and permit higher dosing.150 Additional micelles, including NK105 (paclitaxel), NK911 (doxorubicin), NK012 (SN-38), and NC-6004 (cisplatin), have progressed from phase I to III trials for solid tumors, and several of these have exploited EPR-mediated tumor accumulation with reduced off-target toxicity.151,152Inorganic nanoparticlesInorganic nanoparticles, composed of metals, metal oxides, silica, or semiconductors, offer optical, magnetic, electrical, and catalytic functionalities unavailable to organic carriers. Their degradation-resistant cores provide high stability, large surface-area-to-volume ratios enable dense ligand functionalization, and many exhibit intrinsic responsiveness to light, magnetic fields, or radiation, thus enabling spatiotemporally controlled therapy and integration of diagnostic and therapeutic functions within a single theranostic platform (Table 3).153Table 3 Summary of inorganic nanoparticlesFull size tableClassification and inorganic-specific propertiesFour classes are commonly distinguished. Metal-based nanoparticles (gold, silver, iron) show surface plasmon resonance (SPR), catalytic activity, and magnetic responsiveness. Ceramic nanoparticles (silica, TiO₂, alumina, calcium phosphate) provide structural stability, biocompatibility, and versatile surface chemistry. Semiconductor quantum dots (QDs) demonstrate size-dependent optical and electronic properties from quantum confinement. Hybrid core–shell systems (e.g., gold–silica, magnetic–gold) integrate complementary functions for multimodal imaging and therapy.154,155,156,157,158Beyond the shared size and charge determinants (see Synthetic intro), inorganic particles offer engineered morphology and intrinsic physics that are unavailable to organic carriers. Controlled nucleation and growth yield nanorods, nanocubes, and nanostars with distinct uptakes and photothermal behaviors, because high-aspect ratios and high-curvature geometries enhance membrane interaction and cellular uptake. Gold SPR enables efficient near-infrared (NIR) light-to-heat conversion for photothermal therapy;159,160 superparamagnetic iron oxide provides magnetic resonance imaging (MRI) contrast and under alternating magnetic fields, generates localized heat for magnetic hyperthermia; mesoporous silica offers tunable pores and high surface area for high-capacity loading and controlled release;161,162,163 and semiconductor QDs, including cadmium-based and emerging heavy-metal-free (ZnO, Zr-based) variants, provide size-tunable, photostable, narrow-emission fluorescence with broad excitation, which are well suited to long-term multiplexed and high-resolution bioimaging.164 A key design constraint for inorganic particles is safe elimination, as a hydrodynamic diameter below about 5.5 nm is required for efficient renal clearance and long-term tissue accumulation avoidance.165 Ceramic nanoparticles such as mesoporous silica and hydroxyapatite additionally offer tunable porosity and structural stability for high-capacity loading and bone-related applications. The high surface-area core supports dense functionalization: biocompatible polymers such as PEG improve colloidal stability and circulation by limiting protein adsorption, targeting ligands (antibodies, peptides, small molecules) enable receptor-specific delivery, and stimulus-responsive linkers permit pH-, redox-, or enzyme-triggered release within pathological microenvironments.92,166Clinical validationSeveral inorganic systems have reached the clinic.167 Ferumoxytol, an iron-oxide nanoparticle, is FDA-approved for iron-deficiency anemia and has been explored for MRI and delivery; AuroShell, a silica–gold nanoshell, was developed for NIR photothermal tumor ablation; and NBTXR3, a hafnium-oxide radioenhancer, has been clinically investigated for improving radiotherapy efficacy in solid tumors.168 Beyond these approved products, inorganic platforms are actively pursued as theranostic agents that combine imaging and therapy, with gold and iron-oxide systems integrating photothermal or magnetic-hyperthermia treatment with real-time MRI or optical tracking, thus reflecting their principal translational advantage over purely organic carriers. However, a modality-specific limitation is poor biodegradability: gold, silica, and iron-oxide particles demonstrate slow clearance and prolonged hepatic and splenic retention, raising long-term safety concerns that constrain repeated dosing. This has motivated the development of ultrasmall (sub-8 nm) renal-clearable particles and biodegradable inorganic frameworks designed to be excreted after completion of their diagnostic or therapeutic function.169Conjugate systemsTargeted molecular conjugates, including antibody-, peptide-, and small molecule-drug conjugates, are an important class of precision DDS. Constructed through direct conjugation of therapeutic payloads to targeting ligands, these systems enable selective delivery of highly potent drugs to diseased tissues while minimizing systemic toxicity and off-target exposure. Their modular design supports versatile combinations of targeting molecules, linkers, and payloads, thus facilitating broad clinical applications in precision therapeutics (Table 4).Table 4 Summary of ADCs, PDCs, and SMDCsFull size tableADCsADCs unite the tumor-targeting specificity of a monoclonal antibody with the cytotoxic potency of a small-molecule payload, widening the therapeutic window relative to that of conventional chemotherapy.170 An ADC comprises three engineered elements, the antibody, linker, and payload, whose individual and combined optimization determines efficacy, safety, and pharmacokinetics (Table 4).171,172,173,174Target selection and antibody formatsEffective ADC targets are antigens that are highly and homogeneously expressed on tumor cells but minimally on normal tissue, and ideally internalizing to allow intracellular payload release. The evolution of ADCs is often described in three generations: first-generation constructs (e.g., gemtuzumab ozogamicin) used murine antibodies and unstable linkers with narrow therapeutic windows; second-generation agents adopted humanized antibodies, higher drug loading, and more stable linkers; and third-generation ADCs employed site-specific conjugation, homogeneous drug-antibody ratios, and novel payloads.175 Humanized and fully human IgG1 antibodies predominate owing to their favorable half-life and effector functions, while the internalization and trafficking behavior of the target antigen critically shape payload delivery efficiency.171,176,177 Representative validated targets include HER2 (breast and gastric cancers), CD30 (lymphoma), CD22 and CD19 (leukemia), TROP2 and Nectin-4 (solid tumors), reflecting a broadening from hematologic to solid-tumor indications with advancing antibody engineering and payload chemistry.178,179Linkers, payloads, and drug-antibody ratioThe linker balances plasma stability against efficient intracellular release, and linker chemistry is a primary determinant of ADC efficacy and toxicity.180 Cleavable linkers exploit tumor-relevant conditions, such that acid-labile chemistries release the payload in endo-lysosomes, protease-sensitive (e.g., valine-citrulline) linkers are cleaved by cathepsin B, and disulfide linkers respond to intracellular glutathione;181 in contrast, non-cleavable linkers rely on complete antibody degradation and can offer greater plasma stability. The payload sets the potency of ADCs: microtubule inhibitors such as auristatins (MMAE, MMAF) and maytansinoids (DM1, DM4) historically predominated owing to sub-nanomolar activity, while DNA-damaging agents (calicheamicins and duocarmycins) and, more recently, topoisomerase-I inhibitors (deruxtecan, SN-38), have enabled new approvals and bystander-killing effects in heterogeneous tumors.40 The bystander effect, wherein a membrane-permeable payload released from a targeted cell diffuses to kill neighboring antigen-low cells, is particularly valuable in tumors with heterogeneous antigen expression, and is tunable through payload hydrophobicity and linker cleavability, though it also contributes to off-target toxicity, which must be balanced against efficacy.182,183The drug-antibody ratio (DAR), the average number of payload molecules per antibody, is among the most critical determinants of the therapeutic window. An excessively high DAR increases hydrophobicity, promoting aggregation, faster clearance, and reduced efficacy, whereas a low DAR limits potency.184,185 Traditional conjugation to native lysines or interchain cysteines produces heterogeneous DAR distributions with batch-to-batch variability; site-specific strategies such as engineered cysteines, unnatural amino acids, and enzymatic (e.g., transglutaminase, glycan-based) methods, yield homogeneous, well-defined conjugates with improved stability and reproducibility.186Clinical progress and approved ADCsMultiple ADCs are now integrated into oncology practice. Trastuzumab emtansine (T-DM1), the first ADC approved for solid tumors after it prolonged survival in HER2-positive advanced breast cancer in the EMILIA trial,187 is approved for HER2-positive breast cancer, including the adjuvant setting for residual invasive disease after neoadjuvant therapy.188 Brentuximab vedotin (anti-CD30–MMAE) is approved for relapsed Hodgkin lymphoma and systemic anaplastic large-cell lymphoma,189 and inotuzumab ozogamicin (anti-CD22–calicheamicin) markedly improves remission rates in relapsed/refractory acute lymphoblastic leukemia.190 Trastuzumab deruxtecan has broadened HER2-directed therapy through a high-DAR topoisomerase-I payload with bystander activity, extending benefits even to HER2-low breast cancer, which was previously considered outside HER2-targeted treatment, as shown in the DESTINY-Breast04 trial.191 Sacituzumab govitecan (anti-TROP2–SN-38) and enfortumab vedotin (anti-Nectin-4–MMAE) have similarly expanded ADC use to triple-negative breast and urothelial cancers, underscoring the platform as one of the fastest-growing classes in oncology.191,192 However, a modality-specific limitation is the off-target toxicity from premature payload release and antigen expression on normal tissue, driving the shift toward site-specific conjugation and safer linker–payload chemistries (Section “A unified framework for delivery-system bottlenecks”).Peptide-drug conjugates (PDCs) and small molecule-drug conjugates (SMDCs)Peptide-drug conjugates (PDCs) and small molecule-drug conjugates (SMDCs) are targeted systems in which a bioactive peptide or small-molecule ligand is covalently linked, through a cleavable linker, to a cytotoxic or radioactive payload. Their smaller size relative to antibodies confers deeper tissue penetration, lower immunogenicity, and simpler, lower-cost synthesis, at the cost of a shorter circulatory half-life.193,194,195,196Architecture and mechanismsThree modular elements define these conjugates: a targeting ligand, linker, and payload. In PDCs the ligand is a peptide that recognizes tumor-overexpressed receptors;73 in SMDCs it is a low-molecular-weight ligand such as a folate or PSMA-binding moiety directed at receptors, enzymes, or transporters.74 Cleavable linkers, which are responsive to enzymatic cleavage (cathepsin B), acidic pH, or reductive conditions, selectively release the payload within the target microenvironment, while non-cleavable linkers depend on ligand internalization and processing.185,197 After ligand–receptor binding and internalization, the payload is released to exert its cytotoxic, radiotherapeutic, or imaging effect, with selectivity principally governed by differential receptor expression. Peptide ligands offer high receptor affinity and structural tunability but can be susceptible to proteolytic degradation, thus motivating cyclization, D-amino-acid substitution, and stapling to improve stability; small-molecule ligands such as folate and PSMA inhibitors provide metabolic robustness and defined chemistry.198 Payloads extend beyond cytotoxins to radionuclides for peptide-receptor radionuclide therapy and to fluorophores for image-guided surgery, positioning PDCs and SMDCs as a versatile theranostic bridge between molecular conjugates and nanoparticle carriers. A modality-specific trade-off is that small size, while aiding penetration, causes rapid renal clearance, thus requiring dose and linker optimization for sustained exposure.199Clinical progress and approved PDCs/SMDCsClinical translation is advancing, particularly in oncology-targeted radiopharmaceuticals and imaging. Lutetium-177 DOTATATE (Lutathera), a somatostatin receptor–targeting radioconjugate, was the first radiopharmaceutical of its class approved for gastroenteropancreatic neuroendocrine tumors (GEP-NETs).200 PSMA-targeting radioligand conjugates have since shown marked benefit in metastatic prostate cancer. In fluorescence-guided surgery, the folate-receptor–targeted agent pafolacianine (Cytalux) enables intraoperative visualization of ovarian and lung tumors.201 Multivalent ligand designs are being explored to improve target avidity and selectivity.202 Notably, the GalNAc-conjugated siRNA inclisiran (PCSK9), a small-molecule-ligand conjugate, achieves durable hepatocyte-targeted gene silencing.Biological delivery systemsBiological delivery systems use living cells, bacteria, viruses, or bacteriophages as carriers, exploiting innate targeting, self-amplification, and biological functions, which are difficult to engineer synthetically. A recurring theme is that several of these platforms act simultaneously as a delivery vehicle and therapeutic agent, a duality that shapes both their promise and translational challenges.81,203Cell-mediated drug deliveryCell-based drug delivery exploits the intrinsic biological properties of circulating and tissue-resident cells to enhance therapeutic transport and biodistribution. Living cells possess unique capabilities for systemic circulation, selective homing to pathological sites, and dynamic interactions with the tumor or inflammatory microenvironment. These properties facilitate navigation across biological barriers, prolonged retention at disease sites, and controlled payload release, enhancing therapeutic efficacy while reducing systemic toxicity (Table 5).204,205,206Table 5 Summary of cell-based drug delivery strategiesFull size tableErythrocytes, immune cells, and stem cellsErythrocytes (RBCs) are among the most studied carriers: their abundance, circulatory compatibility, and ~120-day lifespan support sustained release, and their cargo can be loaded internally or onto the surface, providing a platform for nanoparticle hitchhiking that enhances pharmacokinetics and modulates immune responses.207 Neutrophils exploit rapid chemotaxis toward inflamed or malignant tissue along chemokine gradients (e.g., IL-8, CXCL family) to breach barriers such as the inflamed brain; neutrophils carrying drug-loaded liposomes have been used to suppress post-surgical glioma recurrence, with inflammatory signals triggering site-specific cargo release.208,209,210Macrophages and monocytes provide strong phagocytic capacity and pronounced tropism for tumors and inflamed tissue, internalizing nanocarriers and migrating into hypoxic tumor regions that are poorly accessible to passive systems. Mesenchymal stem/stromal cells (MSCs) home to injured or neoplastic sites through tumor-tropism, making them attractive for site-specific delivery of therapeutic proteins, oncolytic viruses, or nanoparticles.211,212,213,214 The tumor-homing of MSCs is driven by chemokine–receptor axes (e.g., CXCL12–CXCR4, CCL2–CCR2) mirroring stromal cell recruitment to sites of tissue remodeling and can be enhanced by genetic modification or membrane engineering.215 A safety consideration specific to MSCs is their potential, once integrated into the tumor stroma, to contribute to tumor-supportive signaling, which motivates careful control of MSC dose, persistence, and functional state. RBC hitchhiking, which involves transient adsorption of nanoparticles onto the erythrocyte surface, exploits long circulation and shear-dependent detachment pulmonary microvasculature to enhance lung-targeted delivery, while erythrocyte membrane–camouflaged nanoparticles transfer RBC stealth proteins (e.g., CD47 “don’t-eat-me” signals) onto synthetic cores to evade phagocytic clearance.216 Across cell types, a shared consideration is preserving carrier viability and controlling premature cargo release, as intracellular loading can compromise cell function whereas surface loading risks off-target detachment.Macrophages and monocytes are especially suited to crossing biological barriers impeding synthetic carriers: their intrinsic transmigration across the BBB and infiltration into hypoxic tumor cores enable delivery to otherwise inaccessible regions, and intracellular-pathogen-mimicking strategies load them with nanoparticles for release upon reaching the target. However, macrophage plasticity presents a significant challenge, as carrier cells may be repolarized by the tumor microenvironment toward an immunosuppressive phenotype, potentially counteracting therapy, underscoring the need for controlling the carrier-cell fate after administration.217 Neutrophils are particularly effective for acute inflammatory and post-surgical contexts because their recruitment is rapid and self-amplifying; drug-loaded neutrophils or neutrophil-membrane–coated nanoparticles inherit adhesion molecules that bind the inflamed endothelium, concentrating the payload at sites of active inflammation. In contrast, stem cell carriers are exploited for their sustained tissue-homing and paracrine secretion: MSCs can be engineered to continuously express therapeutic proteins or to carry oncolytic viruses shielded from neutralizing antibodies until they reach the tumor, effectively acting as protective, self-navigating bioreactors.218Chimeric antigen receptor (CAR) therapyChimeric antigen receptor T-cell (CAR-T) therapy has achieved remarkable efficacy in hematological malignancies,219,220 establishing engineered immune cells as a powerful modality. Its limitations, including complex and expensive ex vivo manufacture, antigen escape, limited solid-tumor penetration, and toxicities like the cytokine release syndrome, have motivated next-generation platforms. CAR-natural killer (CAR-NK) cells mediate cytotoxicity through both CAR-dependent recognition and innate perforin/granzyme mechanisms and can be allogeneically used as an off-the-shelf product without the graft-versus-host risk of allogeneic CAR-T.221 CAR-NK cells avoid the severe cytokine release syndrome and neurotoxicity associated with CAR-T,222 can be sourced from peripheral blood, cord blood, or induced pluripotent stem cells, and retain natural cytotoxicity against tumor cells that downregulate the target antigen, thus mitigating antigen-escape relapse. These features position CAR-NK as a scalable, off-the-shelf alternative, but shorter in vivo persistence and efficient manufacturing remain active challenges. Tumor-infiltrating lymphocyte (TIL) therapy, using lymphocytes isolated directly from tumor tissue and expanded ex vivo, presents a complementary, polyclonal adoptive approach, w has achieved durable responses in metastatic melanoma. A shared limitation across adoptive cell therapies is the difficulty of extending their efficacy from hematological to solid tumors, where activity is blunted by an immunosuppressive microenvironment, physical barriers to infiltration, and antigen heterogeneity, which are problems that delivery engineering (armored CARs, local delivery, and combination with microenvironment-modulating agents) aims to address.223,224 A major frontier is the shift from ex vivo to in vivo CAR generation, wherein delivery systems, chiefly targeted LNPs and viral vectors, directly reprogram T cells within the patient, bypassing expensive ex vivo manufacturing. This convergence of CAR therapy with nanoparticle and viral delivery exemplifies how biological and synthetic platforms are increasingly integrated.208,225Bacteria-mediated drug deliveryBacteria function both as a delivery vehicle and as an intrinsic therapeutic agent. Certain bacteria selectively colonize the hypoxic, immunosuppressive tumor core,226 an environment hostile to conventional therapy, and they can be engineered to sense local cues and produce therapeutic payloads in situ.227 This tumor-targeting behavior arises naturally from bacterial physiology, as anaerobic or facultative-anaerobic strains thrive in the necrotic, poorly perfused tumor interior that excludes most systemically delivered drugs, thus giving bacteria an intrinsic targeting mechanism that synthetic carriers must engineer explicitly. The key features of this modality are summarized in Table 6.Table 6 Summary of bacteria-mediated drug delivery platformsFull size tablePlatforms and synthetic-biology engineeringEarly platforms such as Clostridium novyi-NT selectively germinate within hypoxic tumor cores, producing toxin-mediated necrosis in preclinical models; however, systemic inflammation and poor control over dissemination limited translation; engineered attenuated strains (e.g., Salmonella) have since improved safety. Synthetic biology circuits let bacteria act conditionally: engineered genetic programs sense tumor cues (oxygen, pH, metabolites) and trigger payload production, and quorum-sensing–controlled synchronized lysis enables pulsatile intratumoral release of therapeutics.25,228 Bacteriophage-derived extracellular contractile injection systems (eCIS) act as nanoscale devices that physically inject protein payloads across the plasma membrane, bypassing the endocytosis and endosomal escape required by nanoparticle carriers. These programmable capabilities distinguish bacterial delivery from passive carriers: rather than relying solely on biodistribution, engineered bacteria can compute local signals and actuate payload release, and can be designed to secrete therapeutic proteins, nanobodies, or immunomodulators continuously within the tumor, effectively converting the colonized tumor into an in situ drug factory.Clinical translationBacterial cancer therapies have entered early-phase clinical evaluation, chiefly using attenuated Salmonella and Clostridium strains229 engineered for tumor-selective colonization and payload delivery. A modality-specific limitation is the need for stringent biocontainment: auxotrophy and engineered kill-switches are required to prevent uncontrolled proliferation and systemic infection, and immune clearance can limit repeat dosing.230,231 Beyond oncology, engineered commensal and probiotic bacteria are being developed for the in situ production of therapeutics in the gut, exploiting their natural colonization to locally deliver enzymes, cytokines, or metabolites, an approach that reframes the microbiome itself as a programmable delivery platform.Viral-based deliveryViral vectors exploit the natural ability of viruses to enter cells and deliver genetic material with high efficiency, and remain the dominant platform for in vivo gene therapy.70,232,233 Over decades of engineering, wild-type viruses have been stripped of pathogenic genes and repurposed as delivery shells with tunable tropism, cargo capacity, expression duration, and immunogenicity, making them the benchmark against which non-viral systems such as LNPs are measured. Their principal advantage is delivery efficiency, s viruses have evolved specialized machinery for cell entry and, in some cases, nuclear import; however, their principal liabilities are manufacturing complexity, limited cargo capacity, insertional-mutagenesis risk for integrating vectors, and immunogenicity (Table 7).Table 7 Summary of viral-mediated drug delivery platformsFull size tableVector classes: gene-delivery and oncolyticReplication-deficient vectors dominate gene delivery, whereas adenoviral vectors offer high transduction and a large cargo capacity with transient, non-integrating expression; AAVs provide durable expression with a favorable safety profile but restricted packaging capacity;234,235,236 and lentiviral/retroviral vectors enable stable genomic integration for durable expression, chiefly under ex vivo settings such as CAR-T manufacture. In contrast, replication-competent oncolytic viruses selectively infect and lyse tumor cells and beyond direct oncolysis, release tumor antigens and damage-associated signals that activate dendritic cells and prime adaptive anti-tumor immunity, thus remodeling the immunosuppressive microenvironment.237,238 This dual action, direct tumor lysis plus systemic immune priming, means that oncolytic viruses simultaneously function as a delivery vehicle for immunostimulatory transgenes and as an active therapeutic, and combining them with immune-checkpoint blockade has become a major clinical strategy. Vector choice is dictated by application: transient, high-level expression favors adenovirus for vaccination and oncolysis, whereas durable monogenic correction favors AAV, with lentivirus integration reserved for ex vivo modification where permanent expression is required.Clinical validationViral vectors underpin approved gene therapies, including AAV-based products for inherited retinal dystrophy and spinal muscular atrophy;239 further, oncolytic virotherapy is clinically established with talimogene laherparepvec (T-VEC) for melanoma.237 Modality-specific limitations are pre-existing and include vector-induced anti-capsid immunity, which restricts repeat dosing, and the limited packaging capacity of AAV, thus driving the engineering of novel capsids and immune-evasion strategies.240 Directed evolution and rational design of AAV capsids have yielded variants with enhanced tissue tropism,241 including engineered capsids that cross the blood–brain barrier after systemic administration, substantially expanding the range of treatable CNS and systemic disorders and illustrating how vector engineering directly enables new delivery capabilities.242Phage-based deliveryBacteriophages, viruses that infect bacteria, are non-pathogenic to human cells and offer genetically programmable protein coats, making them versatile scaffolds for targeted delivery and display (Table 8).243Table 8 Summary of phage-mediated drug delivery platformsFull size tableStructure and mechanismThe phage capsid can be genetically engineered to display targeting peptides or antibody fragments, and phage display libraries enable the selection of binders with high affinity to disease-associated receptors for targeted delivery.244 As they self-assemble with defined, monodispersed architecture and tolerate extensive surface modification, phages can carry drugs, nucleic acids, or imaging agents while directing them to specific cells. Two structural classes are commonly exploited: icosahedral phages (e.g., MS2, Qβ) provide enclosed capsids for cargo encapsulation, whereas filamentous phages (e.g., M13, fd) present elongated, high-aspect-ratio scaffolds that are ideal for multivalent peptide display. Both can be produced at scale in bacterial hosts and modified genetically or chemically; further, their bacterial rather than mammalian origin gives a favorable safety profile for human cells while enabling dual antibacterial and delivery functions.Engineered phages for targeted therapyEngineered phages and phage-derived nanomaterials are being developed as antimicrobial and anticancer delivery systems, including phage–polymer nanoassemblies and liposome–phage complexes that combine targeting with enhanced stability and payload capacity.243,245 Reprogramming the host range and capsid chemistry further extends their utility against drug-resistant pathogens, and filamentous phages such as M13 have been used for templating targeted nanomaterials and displaying tumor-homing peptides to be used in cancer imaging and therapy. Because their genomes are readily engineered and they self-amplify in host bacteria, phages offer a low-cost, scalable manufacturing route relative to that of synthetic nanocarriers. In antimicrobial applications, engineered phages are entering clinical evaluation for drug-resistant infections, delivering lytic enzymes or CRISPR payloads that selectively kill target bacteria, while in oncology, phage-displayed tumor-homing peptides guide imaging agents and cytotoxic cargo to malignant tissues.246,247 A modality-specific limitation of this system is rapid immune recognition and reticuloendothelial clearance, which shortens circulatory half-life and motivates surface engineering to prolong exposure.248Complementary and emerging platformsBeyond the principal nanoparticle, conjugate, and biological platforms, several complementary systems address specific delivery needs such as local and sustained release, endogenous activation, biomimetic camouflage, cell-derived transport, externally controlled dosing, and defined protein scaffolds. Although diverse, these platforms share an emphasis on matching the delivery mechanism to a specific anatomical, temporal, or physicochemical requirement rather than on broad systemic distribution (Table 9).Table 9 Summary of additional drug delivery strategiesFull size tableHydrogel-based deliveryHydrogels are three-dimensional hydrophilic polymer networks that absorb and retain large amounts of water or biological fluid while maintaining their structural integrity. Owing to their high biocompatibility, tunable mechanics, and extracellular-matrix–mimicking properties, they are versatile platforms for localized and sustained delivery, particularly where an implantable or injectable depot can concentrate therapy at a defined site.249,250Mechanism and designDrug release from hydrogels is governed by diffusion, swelling-mediated transport, polymer-network degradation, and responsiveness to internal or external stimuli. The interconnected porous architecture allows gradual diffusion of encapsulated therapeutics, while degradation of the network supports sustained, controlled release over days to months. The central design lever is crosslinking density: it sets mechanical strength, swelling behavior, degradation kinetics, and diffusion rate, with highly crosslinked networks, resulting in slow degradation and prolonged release. Stimuli-responsive designs, including thermosensitive systems undergoing sol–gel transition at body temperature, or pH-, enzyme-, and light-responsive networks, enable in situ gelation after injection and triggered release within a target microenvironment,251,252 and injectable hydrogels can co-deliver immunomodulators or biologics for local, sustained immunotherapy.253,254Hydrogels are classified by their crosslinking mechanism and material origin. Physically crosslinked gels, formed by ionic interactions, hydrogen bonding, or hydrophobic association, are reversible and mild to encapsulated biologics, whereas chemically crosslinked gels formed by covalent bonds offer greater mechanical stability and controlled degradation. Natural polymers (alginate, hyaluronic acid, collagen, chitosan) provide biocompatibility and cell-adhesive cues, while synthetic polymers (PEG, poloxamers, poly (acrylic acid)) offer tunable, reproducible properties. This versatility supports diverse formats, from injectable in situ–gelling depots, preformed implants, microgels, and nanogels, each matched to a delivery route, ranging from subcutaneous and intratumoral injection to ophthalmic and transdermal application.255Clinical maturityHydrogel platforms, including alginate- and hydrocolloid-based systems, are among the most clinically translated complementary systems, with approved products for wound dressings, tissue fillers, ophthalmic applications, and localized drug delivery.256,257,258 Injectable and in situ–forming hydrogels are advancing for post-surgical local chemotherapy and sustained ocular and intra-articular delivery.259 A modality-specific limitation of these systems is low encapsulation efficiency and burst release of hydrophobic drugs, which is partly addressed by nanoparticle-composite hydrogels, which combine a hydrophobic nanocarrier reservoir with a hydrophilic gel matrix.Prodrug strategiesProdrugs are inactive molecular derivatives of active agents, which are chemically modified with cleavable promoieties or linkers to improve stability, solubility, and pharmacokinetics,260 and designed to regenerate the active drug upon a defined trigger. Historically, the most successful prodrug applications addressed pharmacokinetic shortcomings such as poor solubility, low oral bioavailability, rapid metabolism, or unpleasant taste, but the concept has evolved into a sophisticated targeting strategy that exploits the distinct biochemical microenvironment of diseased tissue for spatially selective activation. As activation occurs at the molecular level within a target, prodrugs can complement carrier-based targeting: the carrier concentrates the agent in the tissue, while the prodrug trigger ensures that the active species is regenerated only where the disease-specific cue is present.10Mechanism and designProdrug activation exploits endogenous stimuli such as enzymatic hydrolysis, acidic pH,261,262 or redox (glutathione, reactive oxygen species) conditions that are characteristic of diseased tissue, to achieve site-selective drug release. Effective design balances two competing requirements: sufficient stability under physiological conditions to prevent nonspecific activation during circulation, and rapid, efficient conversion at the target site. Common cleavable chemistries include hydrazone and acetal linkages for pH-triggered release, disulfide and thioketal bonds for redox-triggered release, and protease-cleavable peptides for enzyme-selective activation. 263,264 Polymer-prodrug and polyprodrug designs, in which therapeutic monomers are directly polymerized, achieve high and well-defined drug loading with reduced premature leakage relative to that obtained with physically encapsulated systems,265 and stimulus-responsive prodrug nanoassemblies can integrate imaging or combination-therapy functions.266Prodrug design also serves goals beyond targeting. Masking of ionizable or reactive groups improves membrane permeability and oral absorption, while transient modification enhances metabolic stability, and self-immolative linkers ensure a clean release of the parent drug without any residual chemical handles.267 Increasingly, prodrug chemistry is integrated with nanocarrier and conjugate platforms such as prodrug-loaded micelles, polymer-prodrug amphiphiles, and antibody-prodrug conjugates, such that carrier-level targeting and molecular-level activation act cooperatively to widen the therapeutic window.Clinical maturityProdrug strategies are highly clinically mature, with numerous approved agents, including capecitabine (oncology), enalapril (cardiovascular), and oseltamivir (infectious disease), which improve bioavailability, enhance oral absorption, and reduce systemic toxicity.261 A modality-specific limitation of these is inter-individual variability in enzymatic activation, as exemplified by CYP2C19-dependent clopidogrel metabolism, thus motivating pharmacogenomics-guided prodrug design.Biohybrid and synthetic-biology–based deliveryBiohybrid systems combine synthetic carriers with living-derived components to inherit biological functions that are difficult to engineer synthetically, most prominently through cell-membrane coating of nanoparticle cores.268 The central rationale is that decades of chemical effort to make nanoparticles “stealthy” and targeted can be short-circuited by borrowing the plasma membrane, an entire biological interface, with its full complement of proteins and glycans, and transplanting it onto a synthetic core. This top-down biomimicry preserves the tunable physicochemical properties of the synthetic core (size, drug loading, magnetic or photothermal function) while adding the identity and functions of the source cell, and effectively merging the strengths of synthetic and biological carriers.Mechanism and designCell membrane-coated nanoparticles cloak a synthetic core with plasma membranes derived from red blood cells, cancer cells, immune cells, platelets, or stem cells, thereby inheriting the membrane proteins and biological functions of the source cell, including immune evasion, prolonged circulation, and homologous (self-recognition) targeting. Membrane orientation is a critical design factor: right-side-out coating preserves stealth and targeting ligands, whereas engineered inside-out or hybrid membranes can expose intracellular domains for specialized applications such as drug-target screening.269 Related strategies exploit engineered cell surfaces, hybrid membranes fusing two cell sources, and synthetic-biology circuits to add programmable functions.270 The key parameters are membrane source (dictating targeting and immune profile), coating completeness and orientation, and core–membrane compatibility. A modality-specific limitation of this system is the unstable nanoparticle–cell attachment under physiological shear, causing premature detachment and variable loading; batch consistency and biosafety are shared constraints.Application-specific membrane sources illustrate the platform’s versatility: red-blood-cell membranes confer immune evasion and prolonged circulation, an approach first established by coating polymeric nanoparticles with erythrocyte membranes;80 cancer-cell membranes provide homotypic tumor targeting through self-recognition adhesion molecules; platelet membranes direct particles to sites of vascular injury and inflammation; and leukocyte membranes support crossing of activated endothelium. Hybrid membranes fusing two cell sources combine complementary functions; for example, RBC–cancer hybrids unite long circulation with homotypic targeting, expanding the design space beyond any single natural cell type.271,272,273Extracellular vesicle-mediated deliveryExtracellular vesicles (EVs), including exosomes, microvesicles, and apoptotic bodies, are naturally secreted lipid-bilayer nanovesicles mediating intercellular communication by transferring proteins, lipids, and nucleic acids.274,275,276 Their intrinsic biocompatibility, low immunogenicity, and natural targeting capacity make them attractive as both therapeutic agents and engineered delivery carriers.277,278Biology and engineeringEVs can act either as intrinsic bioactive mediators, which transfer functional cargo that reprograms recipient cells,49 or as engineered carriers loaded with therapeutic cargo and functionalized with targeting ligands, as first demonstrated by RVG-targeted exosomes that delivered siRNA across the blood–brain barrier.279 As engineered carriers, they can be loaded with exogenous drugs, nucleic acids, or proteins. Cellular uptake proceeds through membrane fusion or endocytosis, with surface proteins mediating cell-specific tropism inherited from the parent cell. Therapeutic cargo is introduced either through donor-cell engineering (endogenous loading, in which producer cells are modified to package the cargo) or post-isolation methods (electroporation, sonication, or incubation), each trading loading efficiency against vesicle integrity. As EVs reflect the molecular status of their parent cells, they are promising diagnostic biomarkers, and this property is exploited to engineer disease-targeting tropism through parent-cell selection or surface display of targeting ligands.280Performance depends on the interdependent choices of the cell source (defining native tropism and baseline cargo), loading strategy, surface engineering for targeting, and purification method. As these parameters strongly influence one another, source selection and scalable, reproducible isolation are the dominant design considerations, and manufacturing standardization remains the principal bottleneck to translation.281,282Surface engineering strategies further tailor EV tropism and function. Genetic fusion of targeting peptides or antibody fragments to EV membrane proteins (e.g., Lamp2b) redirects vesicles to specific tissues, while click-chemistry and lipid-anchor methods enable post-isolation display of ligands.283 To improve manufacturability, producer cells can be immortalized or grown in bioreactors, and biomimetic “EV-mimetic” nanovesicles, generated by extruding cells through membranes, offer higher yields than those of natural secretion. These approaches aim to convert EVs from an intrinsically variable biological product into a controllable, scalable delivery platform.Clinical maturityEV-based therapeutics remain in the nascent-to-intermediate stages of clinical translation. Most candidates are in preclinical development, but a growing number, chiefly mesenchymal stem-cell–derived EVs, have advanced to Phase I–II trials for indications including inflammatory and degenerative diseases.284 Modality-specific challenges include isolation-method trade-offs (ultracentrifugation, tangential-flow filtration, and size-exclusion chromatography differ in yield, purity, and scalability) and population heterogeneity, which together complicate standardized, GMP-compatible manufacturing. Regulatory frameworks for EV therapeutics are still maturing, and defining critical quality attributes such as identity, potency, and purity of a heterogeneous, cell-derived product, remains an active area of standardization efforts for shaping the pace of clinical translation.Device-assisted drug delivery (DADD)Device-assisted systems provide precise spatial and temporal control over drug release through implantable, wearable, or energy-based platforms,285,286,287 thus extending delivery capabilities beyond what carrier chemistry alone can achieve.Mechanisms and designEnergy-based systems use external physical stimuli such as ultrasound, magnetic fields, light, electric fields, radiofrequency, or heat, to enhance drug penetration or trigger controlled release: ultrasound transiently increases vascular and membrane permeability, while magnetic and photothermal approaches enable localized, on-demand activation. Transdermal device technologies, including iontophoresis and microneedle arrays, non-invasively drive drugs across the skin barrier,288,289,290,291 and electronically powered and microfluidic implants provide programmable, closed-loop dosing.292 The design integrates material-host interactions (biocompatibility, fibrosis), transport and release dynamics, energy/power management, and safety-constrained operation; for implantable and electronic systems, power supply, calibration stability, and long-term reliability are the decisive constraints.293The device platforms span a spectrum of invasiveness and autonomy. Non-invasive wearable systems such as iontophoretic patches, microneedle arrays, and ultrasound applicators enhance transdermal or transmucosal transport without surgery, whereas implantable reservoirs, osmotic pumps, and microchip devices provide long-term, programmable dosing for chronic conditions. The frontier comprises closed-loop systems that integrate biosensing with actuation; for example, in glucose-responsive insulin delivery, continuous measurement of a physiological signal automatically adjusts release, facilitating autonomous, patient-specific therapy.Clinical maturitySeveral device systems are FDA-approved and in routine use, including the SynchroMed II implantable infusion pump for chronic pain and spasticity, closed-loop insulin pumps for diabetes, and iontophoretic transdermal systems such as the fentanyl patient-controlled analgesia device.294 Modality-specific barriers include surgical implantation, external-equipment dependence, high cost, and implant-associated risks such as infection, fibrosis, and foreign-body responses.Microneedle systems represent a particularly active translational area, with dissolving and coated microneedle patches advancing vaccine and biologic delivery, offering pain-free self-administration and thermostable formulation.289 The convergence of drug delivery with digital health, through sensor-integrated, wirelessly controlled devices, is expected to enable personalized, adaptive dosing regimens, even though it introduces additional regulatory and cybersecurity considerations distinguishing device-based delivery from purely chemical platforms.Protein-based carriersProtein-based carriers exploit naturally occurring or engineered proteins as delivery scaffolds, combining defined molecular architecture, biodegradability, and low immunogenicity. The most clinically established example is albumin: nab-paclitaxel (Abraxane), an albumin-bound paclitaxel nanoparticle, eliminates the solvent Cremophor EL, eliminating the need for corticosteroid premedication, and improving the therapeutic index relative to that of solvent-based paclitaxel.295 The long circulatory half-life, FcRn-mediated recycling, and gp60/SPARC pathway-mediated passive tumor accumulation of albumin make it a versatile carrier for hydrophobic drugs.296Beyond albumin, self-assembling protein cages provide monodispersed, genetically tunable architectures. Ferritin nanocages offer a hollow, thermostable shell297 whose interior can encapsulate cargo and whose exterior can be genetically fused with targeting ligands, while virus-like particles (VLPs), non-infectious capsids assembled from viral structural proteins without genetic material, present highly ordered, immunogenic surfaces that are useful for vaccine antigen display and targeted delivery. Their precise, atomically defined structure and ease of genetic engineering position protein-based carriers as a bridge between molecular conjugates and nanoparticle systems, even though scalable manufacture and control of immunogenicity remain key translational considerations. Engineered protein scaffolds also permit precise spatial control over ligand valency and payload stoichiometry, which is difficult to achieve with lipid or polymeric carriers, as each subunit occupies a genetically defined position. This atomic-level definition, combined with intrinsic biodegradability into amino acids, gives protein-based carriers a distinctive translational profile with predictable composition and metabolism, but a dependence on recombinant expression and self-assembly fidelity, which shapes their manufacturability. Ongoing work on de novo–designed protein nanocages and on reducing anti-carrier immune responses aims to broaden their use from vaccine antigen display to systemic therapeutic delivery.Collectively, these platforms illustrate the remarkable diversity of contemporary DDS, ranging from synthetic nanoparticles and molecular conjugates to living cellular and viral carriers. Representative FDA-approved and clinical-stage nanotherapeutics across these modalities are summarized in Table 10. Despite differences in their composition, payload, and targeting strategy, all delivery systems ultimately face the same sequence of biological barriers after administration. Therefore, their clinical performance depends on platform-specific design as well as on their ability to overcome common challenges in circulation, tissue distribution, cellular uptake, and intracellular trafficking.298 These shared mechanistic constraints form the basis of the multiscale framework discussed in the following section.Table 10 Current status of FDA-approved and clinical-stage nanotherapeuticsFull size tableMultiscale biological barriers: mechanistic frameworkDelivery efficiency depends as much on the biological barriers that a carrier encounters after administration as on the design of the carrier itself. These barriers arise in a fixed order. A carrier first circulates, then enters and distributes within the tissue, and finally undergoes cellular uptake and intracellular processing.18 At each of these steps a part of the dose is lost, and the fraction that finally reaches its site of action in the functional form is reduced accordingly.37,299,300The barriers are not independent, and a change that helps at one step can shift the outcome at another. Longer circulation raises exposure at the target and increases uptake by clearance organs; further, greater cellular uptake does not guarantee endosomal escape or correct intracellular localization. Owing to these dependencies, optimizing any single parameter is rarely enough, thus necessitating a framework covering the entire sequence.301,302A multiscale framework addresses delivery at three levels, systemic, tissue, and cellular, allowing the step limiting overall efficiency to be identified and platforms to be compared on the same terms. Nanoparticles, conjugates, viral vectors, and cell-based systems work through different mechanisms, but meet the same broad classes of constraint.11,18 This shared set of constraints yields a common basis for judging performance and recognizing where limitations are held in common (Fig. 3).Fig. 3Full size imageMultiscale biological barriers governing drug delivery from systemic circulation to functional therapeutic outcome. Therapeutic delivery is influenced by a series of biological barriers that operate sequentially from systemic circulation to intracellular target engagement. This figure organizes these processes into four interconnected levels: systemic transport, tissue distribution, cellular trafficking, and functional therapeutic outcome. The systemic level includes biological processes that regulate circulation time, biodistribution, and initial interactions between delivery systems and the host. Protein corona formation, immune recognition, and clearance by the mononuclear phagocyte system, liver, spleen, and kidneys influence carrier stability and determine the fraction of administered therapeutics that remains available for delivery. These processes are affected by particle size, surface chemistry, composition, and other physicochemical properties of the delivery system. After systemic transport, therapeutic agents encounter tissue-specific barriers that restrict extravasation and local distribution. Vascular permeability, endothelial transport, extracellular matrix architecture, interstitial pressure, and tissue organization collectively determine the extent of tissue penetration. Additional organ-specific barriers, including the blood–brain barrier, dense tumor stroma, and fibrotic extracellular matrices, further limit access to target cells and contribute to marked differences in delivery efficiency among disease settings. Successful tissue delivery does not necessarily result in productive intracellular transport. Following cellular uptake, therapeutic carriers are directed through multiple endocytic pathways that may lead to recycling, endosomal maturation, lysosomal degradation, or cytosolic release. For many intracellular therapeutics, particularly nucleic acid–based medicines, endosomal escape remains a major rate-limiting process because only a small proportion of internalized cargo reaches the cytosol in a biologically active form. The final stage considers functional delivery rather than cellular uptake alone. Therapeutic activity depends on successful intracellular trafficking, release of intact cargo, and interaction with the intended molecular target. The lower left panel summarizes key factors that influence delivery performance across biological scales, including physicochemical properties, biological interactions, disease state and patient variability, administration route and dosing regimen, and payload properties. These variables interact across multiple levels of biological organization and collectively determine the therapeutic outcome of drug delivery systems. The lower right panel presents representative assay for evaluating productive intracellular delivery, including Gal8 recruitment, split-luciferase reporters, FRET-based approaches, and functional reporter systems. These methods serve as complementary tools for assessing delivery efficiency by distinguishing functional target engagement from total cellular uptake. Created with BioRender.comSystemic barriers: circulation, protein interactions, and clearanceAs soon as a carrier enters the blood it encounters plasma proteins and circulating cells, and proteins adsorb onto its surface within seconds to form a protein corona.7 The composition of this corona reflects surface charge, hydrophobicity, and size; it also varies with the plasma of different individuals and disease states.303,304 Once formed, the corona gives the carrier its biological identity, covering the targeting ligands and creating new interactions that alter both specificity and immune recognition.305,306 As in vitro systems rarely reproduce this process, corona formation is one of the main reasons for the divergence of in vitro and in vivo.307,308Clearance places an additional limit on systemic exposure. The liver, spleen, and kidney remove carriers according to their size and surface properties, and for particulate systems, uptake by Kupffer cells and splenic macrophages dominates.309 PEGylation decreases protein adsorption and lengthens circulation, but this gain is not durable. PEG can be displaced over time, and repeated dosing may induce anti-PEG antibodies that accelerate blood clearance.310,311 As longer circulation raises both target exposure and off-target accumulation, circulation time serves as a surrogate for delivery efficiency rather than a direct measure.312 Nucleic-acid and viral therapeutics face the same systemic constraints.67,69Tissue-level barriers: vascular transport and microenvironmentOnce out of circulation, carriers cross vascular barriers whose permeability differs from organ to organ, then travel through the tissue by diffusion and convection. Both modes of transport are often impaired in diseases, with raised interstitial pressure and abnormal vasculature preventing uniform distribution.313 The extracellular matrix adds resistance through its collagen and glycosaminoglycan networks and through electrostatic and hydrophobic binding.314,315 When binding is strong, carriers remain near the point of entry and penetrate poorly, a limitation known as the binding-site barrier; when binding is weak, distribution broadens but clearance speeds up.316 As the endothelium and matrix are heterogeneous, delivery becomes spatially uneven, and this unevenness contributes to a variable therapeutic response.317,318Organ-specific barriers: BBB and tumor microenvironmentThe BBB limits entry into the central nervous system through endothelial tight junctions, a low rate of transcytosis, and active efflux.30,319 Carriers usually cross by receptor-mediated transcytosis, but binding a receptor does not ensure transport. Receptors can recycle without carrying their cargo across, so affinity has to be tuned for efficient release,320,321 and receptor expression itself varies by brain region and disease state.322 In solid tumors the vasculature is permeable but perfused unevenly, and dense matrix together with high interstitial pressure works against inward transport. Uptake by perivascular cells then sets up concentration gradients that keep carriers from penetrating deep into the tissue.323,324,325Cellular barriers: uptake and intracellular traffickingEffective delivery requires both uptake and correct routing inside the cell. Carriers enter via clathrin-mediated endocytosis, caveolae-mediated uptake, and macropinocytosis, and the balance among these routes depends on the carrier properties, cell type, and cell state.326 The routes run in parallel and shift as the cell condition changes.327,328 The route of entry shapes the following process, as some pathways partly bypass lysosomal processing whereas others lead straight to degradation; therefore, uptake efficiency on its own is a poor performance predictor.329 A surface modification that raises internalization need not improve functional delivery if the material remains trapped in vesicles or is broken down.67,330,331Endosomal escape and rate-limiting stepsAfter uptake, carriers move into endosomes, and the cargo held there can be degraded or recycled before it reaches the cytosol or nucleus. This loss is especially severe for nucleic acids and the proteins that act inside the cell.332 Endosomal escape is widely considered the rate-limiting step, because efficient uptake often coincides with the release of only a small fraction of the cargo, explaining why uptake-based measurements overstate functional delivery.66,333 Direct imaging of LNPs has placed this escape at only about 1–2% of the internalized siRNA.66,99 The proposed escape mechanisms include pH-driven destabilization of the endosomal membrane during acidification and direct membrane perturbation or fusion, and these often act together.42,334 Escape remains inefficient and varies with cell type. Greater membrane disruption improves release but also increases toxicity; therefore, the balance between release and tolerability remains a central problem.117,335Productive delivery and functional evaluationProductive delivery, the fraction of material that reaches its site of action in a functional form, separates physical accumulation from biological effect, as high uptake does not by itself produce the intended outcome.336 Material is lost through lysosomal degradation, non-productive retention, or release in an inactive form, therefore, performance rests on both transport efficiency and payload functionality, and two systems with similar accumulation can behave very differently. Therefore, uptake-based measurements are insufficient, and functional assays are needed to confirm that the cargo arrives at the right compartment in an active state.337Mechanistic assays probe individual steps missed by bulk measurements. Galectin-8 (Gal8) recruitment reflects endosomal disruption, and split-luciferase reporters provide a readout of delivery into the cytosol, with downstream measurements of gene or protein activity completing the picture.338,339 As reported efficiencies shift with assay design, cell type, and conditions, standardized and reproducible evaluation of function is essential before a meaningful comparison of systems.116Integration across delivery systemsThe same barriers apply across platforms, even though their relative weight shifts with the payload and disease setting. Intracellular trafficking tends to limit nucleic acids, whereas distribution and clearance barriers are dominant for small molecules. A multiscale framework provides a consistent structure for comparing systems with different composition, separating shared limitations from platform-specific ones, and shifting emphasis from individual platforms to the biological constraints that finally determine functional delivery.340Clinical applications across disease contextsSection “Multiscale biological barriers: mechanistic framework” described delivery barriers, namely systemic, tissue, cellular, and intracellular barriers, as a general mechanistic framework. However, their relative importance differs sharply with disease, because tissue architecture, vascularization, and the intended molecular target vary across pathologies. Rather than re-deriving these barriers, this section identifies the barriers that dominates in each of five representative disease areas, and then examines the subsequent delivery challenges and delivery systems best suited to overcome them, with an emphasis on clinically approved and late-stage examples. The governing constraint differs in each case, including interstitial transport and pressure in solid tumors, the blood–brain barrier in neurodegenerative disease, flow and retention in cardiovascular disease, hepatic tropism in metabolic disorders, and immune recognition in autoimmune disease. As delivery performance can only be judged against a specific biological context, a disease-oriented comparison is more informative than an evaluation of platforms in isolation. Figure 4 summarizes this disease-oriented framework by illustrating how the dominant biological barrier in each therapeutic area defines the principal delivery challenge and guides the selection of appropriate delivery strategies.Fig. 4Full size imageDisease-specific biological barriers and delivery challenges across therapeutic areas. Biological barriers differ substantially among diseases and represent one of the primary determinants of delivery performance. This figure compares the major pathological features that influence therapeutic transport, the corresponding delivery challenges, representative drug delivery systems (DDS), and key design considerations across five major disease groups: oncology, neurodegenerative diseases, cardiovascular diseases, metabolic disorders, and autoimmune diseases. Each disease presents a distinct biological environment that influences the selection of delivery strategies. Solid tumors are characterized by dense extracellular matrices, elevated interstitial pressure, abnormal vasculature, and heterogeneous target expression, all of which restrict tissue penetration and homogeneous drug distribution. In neurodegenerative disorders, the blood–brain barrier remains the principal obstacle to therapeutic transport into the central nervous system. Cardiovascular diseases present different constraints, including rapid blood flow, limited vascular retention, and restricted accumulation within injured tissues. Metabolic disorders are frequently associated with preferential hepatic uptake, creating additional challenges for extrahepatic delivery, whereas autoimmune diseases require selective regulation of immune cell populations while minimizing systemic immunosuppression. These biological characteristics give rise to different delivery requirements. Depending on the disease, effective delivery may depend on improved tissue penetration, efficient barrier crossing, prolonged local retention, intracellular transport, or selective immune modulation. Representative delivery platforms, including lipid nanoparticles, antibody–drug conjugates, viral vectors, extracellular vesicles, engineered immune cells, hydrogel-based systems, and ligand-directed carriers, are shown as examples of approaches currently applied to address these challenges. The final column summarizes practical design priorities for each disease setting. Although individual delivery platforms differ in composition and mechanism, platform selection is ultimately guided by the dominant biological constraints associated with each disease. Delivery strategies therefore evolve in response to disease biology rather than being universally applicable across therapeutic indications. Created with BioRender.comOncologyOncology has been the primary driver of drug-delivery innovation, and most approved nanomedicines have been developed for cancer.55,341 The rationale is that as cytotoxic agents have narrow therapeutic windows, improving their accumulation in tumor tissues relative to that in healthy tissue directly improves the balance of efficacy and toxicity. Over decades, platforms across every major modality, ranging from liposomes, polymeric micelles, antibody–drug conjugates, to cell therapies have been clinically established; however, the central quantitative problem of inefficient and heterogeneous tumor delivery remains only partially solved, which is why oncology continues to define the frontier of the field.Dominant barrier and delivery challengeOf the tissue-level barriers described in section “Multiscale biological barriers: mechanistic framework”, the solid-tumor microenvironment is the most restrictive. The defining problem is not reaching the tumor vasculature but rather distributing uniformly once there. This is because a dense extracellular matrix and elevated interstitial fluid pressure suppress convective transport, thus confining systemically administered carriers to perivascular regions and leaving the hypoxic, poorly perfused areas untreated.323 Elevated interstitial fluid pressure is a recognized and long-standing obstacle to transcapillary drug transport in tumors.342 The EPR effect, long considered the basis of tumor-targeted delivery, is now recognized as highly heterogeneous in human tumors and frequently insufficient to produce uniform accumulation.62,343 The scale of the problem is captured by a widely cited analysis reporting that a median of roughly 0.7% of an injected nanoparticle dose reaches tumor tissue,35 a figure indicating that tissue-level transport imposes limits as strong as those of as systemic clearance.18 Active targeting does not resolve this; high-affinity ligands can create a binding-site barrier that retains particles near vessels and further reduces deep penetration, revealing a trade-off between binding avidity and interstitial diffusion.35,316,344 Intravital and whole-tumor imaging studies reinforce this picture,345 demonstrating that nanoparticles extravasate but predominantly accumulate in perivascular regions and move slowly through the interstitium, and that regions with high cellular density or compact matrix consistently receive less drug than that in more permissive microenvironments even with a constant particle design. Competitive uptake compounds the problem: cells adjacent to functional vessels internalize a disproportionate share of the dose, while stromal populations such as cancer-associated fibroblasts and tumor-associated macrophages, sequester nanoparticles before they reach malignant cells;346,347,348 however, this same macrophage avidity has been therapeutically redirected to reprogram the immunosuppressive microenvironment. Once internalized, endosomal entrapment and lysosomal degradation further reduce the fraction of intact cargo reaching its subcellular target. Therefore, effective delivery in oncology must simultaneously contend with vascular access, interstitial transport, competitive cellular capture, and intracellular trafficking.349 Strategies that actively drive nanoparticles deeper into avascular tumor regions, for example by conjugating them with tumor-infiltrating immune cells, have been developed to specifically overcome this transport limitation.350,351Applicable delivery systemsLiposomes were the first nanocarriers developed to overcome systemic barriers such as rapid clearance, drug instability, and off-target toxicity, by encapsulating cargo in a lipid bilayer and, when PEGylated, evading the MPS to prolong circulation. The first FDA-approved nanomedicine, PEGylated liposomal doxorubicin (Doxil), exemplifies the resulting clinical gain as it markedly reduces the dose-limiting cardiotoxicity of free doxorubicin while preserving antitumor activity.56 Its limitations also illustrate the intratumoral barrier: at 80–100 nm, liposomes extravasate but penetrate poorly against high interstitial pressure, owing to which, accumulation is often confined to the tumor periphery rather than the hypoxic core. The liposomal platform has nonetheless matured across multiple products, including liposomal daunorubicin, vincristine, and irinotecan; further, the CPX-351 fixed-ratio cytarabine–daunorubicin formulation approved for acute myeloid leukemia after a phase 3 trial showed improved median overall survival over standard cytarabine-plus-daunorubicin chemotherapy (9.6 versus 6.0 months),352 demonstrating that controlled pharmacokinetics and reduced toxicity translate into a clinical benefit even when deep penetration is imperfect. Active-targeting liposomes bearing antibodies or peptides against tumor antigens have also been pursued to add cellular selectivity, and although antigen-targeted liposomes have been slower to gain approval than previously anticipated, they illustrate the general progression of the field from purely passive to increasingly active designs.353Polymeric nanoparticles and micelles address a different constraint, namely drug solubility and controlled release. Genexol-PM, a paclitaxel-loaded polymeric micelle, eliminates the toxic surfactant Cremophor EL, thus improving tolerability and permitting higher doses in advanced non-small-cell lung and breast cancer, with an objective response rate close to 38% when combined with cisplatin in lung cancer.354,355 To overcome the same interstitial barrier that limits liposomes, next-generation polymeric systems adopt three strategies: ultra-small (