Introduction
The lung’s extensive epithelial surface and the thin alveolar barrier make pulmonary administration an attractive strategy for both localized and systemic drug delivery. This route can significantly reduce diffusion distances while avoiding degradation in the gastrointestinal tract and first-pass metabolism in the liver[
1,
2]. However, physiological challenges such as airway obstruction, localized hypoventilation, and alterations in tissue structure can result in uneven aerosol deposition, while rapid transepithelial transport may increase systemic exposure[
1,
2]. Moreover, the therapeutic targets differ considerably across various respiratory diseases. For instance, antimicrobial and antitumor therapies aim to eliminate actively dividing cells, whereas addressing inflammatory conditions focuses on restoring immune balance, and treatments for fibrotic or degenerative diseases concentrate on maintaining or regenerating functional tissue. As a result, the success of treatment is affected by both the amount of medication that reaches the specific lung region and the suitability of the treatment method for the underlying condition.
Once deposited, inhaled therapeutics encounter various clearance mechanisms in the conducting airways and alveoli. Within the upper airway, mucociliary transport clears particles from the proximal airways, whereas further down in the lungs, immune cells such as alveolar macrophages capture foreign substances through the process of phagocytosis[
3,
4]. Pulmonary surfactant can naturally improve drug dispersion but may also alter membrane interactions, aggregation, and macrophage uptake[
5]. Furthermore, disease drastically changes these barriers. Illnesses such as cystic fibrosis result in the production of thick mucus, whereas other conditions may lead to inflammation, obstructions, or the accumulation of bacteria[
1,
6,
7]. Due to the diversity of these physical changes, a delivery system that performs well in healthy lungs may behave differently in diseased tissue (Figure 1).
Biological treatments, such as specific bacteria, mammalian cells, or microalgae, could help overcome the persistent delivery barriers. Unlike conventional drugs, these living therapeutics can actively respond to their environment. Depending on the disease, they can sense their surroundings, move to specific sites, and release therapeutic agents locally where they are needed most. Similarly, therapeutic viruses and bacteriophages can recognize specific cellular receptors, allowing them to deliver genetic material straight into target cells. Mammalian extracellular vesicles (EVs), outer membrane vesicles (OMVs) and cell membrane-derived vesicles can protect macromolecular cargo and retain surface-mediated interactions[
8]. These properties are accompanied by platform-specific liabilities, including phenotypic instability after administration, immunogenicity and manufacturing variability[
9–
12].
Most studies of biogenic therapeutics evaluate one platform at a time, which makes cross-platform comparison difficult.
Recent reviews have discussed inhaled biologics, living therapeutics, pulmonary drug delivery, EVs and cell-derived carriers. However, most of them focus on one type of biological system or one part of the delivery process, which makes cross-platform comparison difficult (Table 1). The main value of this review is that it brings living and bio-derived systems together within a pulmonary-specific framework. It covers bacteria, bacteriophages, therapeutic viruses, mammalian cells, microalgae, EVs, bacterial OMVs, and other membrane-derived structures. Rather than only listing their applications, we examine how these systems can be engineered at different stages. These stages include biological programming, cargo and biointerface engineering, preservation, and pulmonary delivery. We also discuss how the biological features of each platform affect its engineering needs, aerosol performance, safety and clinical translation. This framework connects platform design with the specific barriers of pulmonary delivery. Therefore, this review contains four parts: (1) compares pulmonary biogenic systems along three axes: biological identity, mechanism of action and final product configuration; (2) classifies the platforms and their principal therapeutic mechanisms; (3) examines engineering strategies that control biological function, cargo, interfaces, formulation and aerosol delivery; and (4) discusses potency, safety, manufacturing, and the evidence required for clinical translation.
Relevant literature was identified through keyword searches on PubMed using pulmonary delivery terms, including lung, airway, inhalation, aerosol, and nebulization, with the search period ranging from 2006 to April 2026, using pulmonary delivery terms, including lung, airway, inhalation, aerosol, and nebulization. Studies were prioritized according to pulmonary relevance, administration route, mechanistic contribution, engineering detail, translational maturity, and recency.
Preclinical studies use several routes to deliver biological systems to the lung. These include intranasal or intratracheal instillation, intratracheal spraying, bronchoscopic delivery, and aerosol inhalation. These routes are not equivalent. However, current studies use different animal models, devices, doses, and procedures. Therefore, this review does not compare each delivery route in detail. Instead, the studies are organized by biological platform and engineering purpose. The reported route is specified when it affects the interpretation of the results. Evidence from direct airway instillation is considered evidence of pulmonary activity, but not proof of aerosol inhalation. Inhalability is discussed only when aerosol generation, biological activity after aerosolization or delivery through an inhalation device has been evaluated.
Categories of pulmonary biogenic therapeutic systems
Scope and mechanistic classification framework
Biogenic therapeutics are first divided into living organisms and acellular derivatives because autonomous metabolism, replication and environmental responsiveness fundamentally affect potency and safety. Within each group, the platforms are further distinguished by their primary mechanism of action and final product configuration.
Living-organism therapeutic platforms
Living therapeutic platforms include bacteria, microalgae, and mammalian cells that remain viable and functional after administration. Depending on the platform, these cells can sense their environment, move actively, produce therapeutic effectors, perform phagocytosis, release paracrine signals, or support tissue regeneration. Such capabilities can provide sustained treatment that adapts to local conditions. However, therapeutic efficacy depends on maintaining cell viability and phenotype, as well as appropriate interactions with the diseased lung microenvironment. Figure 2 summarizes the mechanisms, advantages, and platform-specific limitations of these living systems.
Pulmonary bacterial therapeutics
Functional basis and classification of viable bacterial therapeutics
Pulmonary bacterial therapeutics are viable strains administered to prevent or treat lung disease through native or engineered biological functions. Depending on the strain, these functions may include competition with resident microorganisms, metabolite secretion, bacterial predation, or local production of proteins and antimicrobial factors[
13–
15].
After pulmonary deposition, therapeutic bacteria adjust their transcriptional and metabolic activity in response to oxygen tension, nutrient availability, inflammatory signals, and competing organisms. Pulmonary bacterial therapeutics can therefore be classified as host-directed, pathogen-directed or dual-action according to their principal mechanisms.
Host-directed, pathogen-directed, and dual action
Host-directed bacterial therapeutics modify inflammatory, epithelial or repair pathways without requiring direct killing of the disease-causing organism. Their effects can be mediated by metabolites, cell-envelope signals, and ecological interactions that alter host-microbe signaling. In mouse models of bronchopulmonary dysplasia and chronic obstructive pulmonary disease, an inhaled three-strain lactobacilli product reduced neutrophil influx and inflammatory mediators while improving lung structure and function. Mechanistic experiments linked these effects to suppression of the matrix metalloproteinase-9/proline-glycine-proline neutrophilic pathway[
16]. A second murine study used intranasal
Lactococcus lactis that secreted heme oxygenase-1. Treatment increased pulmonary heme oxygenase-1, reduced elastase-induced alveolar destruction and preserved respiratory function, consistent with antioxidant and anti-inflammatory activity[
17].
Pathogen-directed bacterial therapeutics act by directly killing target microbes or dismantling protective biofilm structures. Repeated intranasal administration of predatory strains (
Bdellovibrio bacteriovorus and
Micavibrio aeruginosavorus) reduced
Klebsiella pneumoniae burden in rat lungs by over three log
10 colony-forming units (CFU), accompanied by transient inflammation and complete clearance of the predators by day 10[
18]. Similarly, an engineered
Mycoplasma pneumoniae strain expressing a bactericidal factor and biofilm-degrading enzymes suppressed acute
Pseudomonas aeruginosa lung infection and disrupted biofilms on endotracheal tubes[
19].
Dual-action bacterial therapeutics combine distinct host-directed pathways with direct pathogen-directed killing[
13–
15]. However, demonstrating both mechanisms requires evidence that each function contributes to the observed therapeutic effect. For example, reduced pulmonary inflammation following bacterial clearance does not alone demonstrate host-directed immunomodulation because it may result entirely from one of the two functions. A dual-action designation is therefore most convincing when experiments distinguish direct antimicrobial activity from host regulation, such as through loss-of-function or matched single-function controls.
Sustained local activity through responsive effector production
A central therapeutic strength of viable bacteria is their ability to integrate local persistence, environmental sensing and continued effector production within a single platform. When the bacterial chassis remains functional, repeated synthesis can replenish proteins or enzymes that would otherwise be diluted, degraded or cleared after administration[
13–
15]. This has the potential to extend the duration of treatment without requiring continuous delivery of the isolated effector.
Some strain-specific properties can also provide additional benefits. Motility and adhesion may increase contact with mucus-associated microbial communities or biofilms, while ecological competition may restrict the growth of pathogenic organisms. Predatory bacteria and living antimicrobial producers can also attack pathogens through mechanisms that differ from those of conventional antibiotics[
18,
19]. These attributes are valuable against drug-resistant or spatially protected infections. However, their contribution depends on the strain, disease environment, and preservation of bacterial function after pulmonary delivery.
Functional variability, persistence, and biocontainment limitations
The same dynamic properties that support sustained activity also complicate the control of bacterial therapeutics. Changes in growth state, nutrient availability and environmental stress can alter effector production after administration. Consequently, CFU can quantify viable recovery but may not accurately represent mechanism-linked potency in the lung.
Genetic instability presents an additional concern. Mutations or selection for less metabolically burdensome phenotypes may preserve viability while reducing the intended therapeutic function[
9,
13,
14,
20]. Furthermore, persistence can also prolong inflammatory stimulation, increase shedding, and create opportunities for horizontal gene transfer. Moreover, cell-envelope components can activate innate immunity, while bacterial lysis may release additional pro-inflammatory material.
In addition, host responses introduce a further source of variability. Rapid immune clearance may remove the therapeutic bacteria before sufficient activity is achieved. Conversely, inadequate immune containment may allow excessive proliferation, particularly in immunocompromised patients. Development therefore requires linked specifications for strain identity, viability, mechanism-specific potency, persistence, clearance, and biocontainment. Preclinical studies should measure both therapeutic efficacy and the recovery, localization, and functional state of the administered bacteria.
Microalgal therapeutics
Motile microalgae as pulmonary delivery platforms
Pulmonary microalgal therapeutics use viable, motile microorganisms primarily as active carriers for local cargo delivery[
21–
23]. Current pulmonary studies have focused mainly on flagellated green microalgae, particularly
Chlamydomonas reinhardtii and the smaller
picoeukaryote Micromonas pusilla. Flagellar propulsion enables these organisms to move through airway liquid and mucus after deposition, potentially redistributing attached cargo beyond its initial deposition site.
Some photosynthetic microalgae can also generate oxygen or other biological products under suitable illumination. However, pulmonary evidence for this function remains sparse, and delivering sufficient light to deep lung regions presents substantial translational challenges. Motility-enabled cargo transport should therefore be regarded as the principal current application, with light-dependent bioproduct generation representing an earlier and less established direction.
Motility-enabled cargo transport
Motility-enabled systems use microalgae to transport and redistribute an associated therapeutic payload after pulmonary administration. In a murine model of
Pseudomonas aeruginosa pneumonia,
C. reinhardtii carrying ciprofloxacin-loaded nanoparticles moved through airway fluid and distributed the antimicrobial cargo within the lung. Treatment reduced bacterial burden and improved survival compared with non-therapeutic controls[
24].
A subsequent study used the smaller
M. pusilla to transport platelet-membrane-coated nanoparticles containing vancomycin in methicillin-resistant Staphylococcus aureus pneumonia. The microalgal carriers retained substantial viability and swimming activity after nebulization. Inhaled treatment improved pulmonary bacterial clearance and survival in mice[
25].
Light-dependent oxygen generation represents a distinct but less developed application. In a bleomycin-induced mouse model of pulmonary fibrosis, inhaled
C. reinhardtii-based microrobots generated oxygen following external near-infrared illumination. The reported oxygen-generation rate was 0.298 ± 0.005 mg·L
−1·min
−1. Treatment increased systemic oxygen saturation, reduced pulmonary hypoxia-inducible factor-1α expression and improved functional outcomes[
26]. Although these findings establish proof of concept, current evidence remains limited to short-term small-animal studies.
Therapeutic value of post-deposition transport through airway mucus
Microalgal motility provides active transport after pulmonary deposition, distinguishing these systems from passive vesicles and synthetic particles. Flagellar propulsion may redistribute attached cargo through airway liquid and mucus, thereby increasing contact with spatially dispersed pathogens or pulmonary lesions[
22–
25]. This ability is useful when mucus, airway geometry, or localized infection restricts the distribution of freely administered therapeutics.
The microalgal chassis can also carry chemically and biologically diverse payloads without requiring microbial colonization or mammalian-cell migration. Protected nanoparticles can reduce direct exposure of the organism to toxic cargo, while partial surface loading can preserve the flagella needed for propulsion. The therapeutic value of these systems therefore lies in coupling an independently optimized payload with a self-propelled biological carrier[
21,
26].
Constraints on mobility, pulmonary clearance, and light-dependent activity
The performance of a motility-driven microalgal system depends on preserving viable and actively swimming cells throughout manufacture, storage, aerosolization, and pulmonary administration. Surface conjugation can damage flagella, increase hydrodynamic drag or reduce cargo retention. Excessive surface loading may therefore increase the nominal payload while reducing the movement needed to distribute it.
Cell size, physiological state and culture conditions can also affect stress tolerance and macrophage-mediated clearance[
25,
27,
28]. Measurements of viability alone are insufficient because living, but poorly motile cells may provide little transport benefit. Functional evaluation should therefore combine viable-cell recovery with swimming velocity, cargo retention, and therapeutic activity after aerosolization.
The translational relevance of microalgal motility also remains uncertain at human anatomical scales. Existing studies provide limited information on whether the organisms can travel sufficient distances within human airway mucus before immobilization or immune clearance. Repeated-dose immunogenicity, persistence, algal biomass clearance, and environmental shedding also require systematic investigation.
Light-dependent systems face additional and potentially more restrictive limitations. Their activity depends on delivering sufficient light to microalgae located within complex and optically heterogeneous lung tissue. Results obtained in small animals may not predict illumination efficiency in humans because of differences in thoracic dimensions and optical path length. Increasing light intensity or exposure time may cause local heating or phototoxicity, while excessive oxygen production could contribute to oxidative tissue injury. Light-assisted microalgal therapy therefore requires direct measurements of intrapulmonary light exposure, temperature, and oxygen distribution, together with comparison against simpler methods of pulmonary oxygen supplementation[
24–
26].
Mammalian cell therapeutics
Functional classes of pulmonary mammalian cell therapeutics
Pulmonary mammalian cell therapeutics use viable cells to regulate tissue injury, remove pathological material or replace damaged epithelial populations. These products can be organized into three broad classes according to their predominant intended function: paracrine regulation, phagocytic activity, or regenerative cell replacement.
Mesenchymal stromal cells (MSCs) primarily act by releasing soluble mediators and EVs that influence epithelial-barrier function, inflammation, and tissue repair. Macrophages instead combine phagocytosis with immune and metabolic regulation, whereas airway basal cells and alveolar progenitors can engraft, self-renew, and generate differentiated epithelial cells[
29–
33].
Paracrine, phagocytic, and regenerative mechanisms
MSCs regulate epithelial and immune-cell behavior through the mediators they release. In a murine model of endotoxin-induced acute lung injury, intratracheally administered bone-marrow-derived MSCs reduced alveolar edema, vascular permeability, and inflammatory-mediator production while improving survival[
34]. This study supports a predominantly paracrine mechanism because durable epithelial replacement was not required to produce the observed effects. However, the MSC secretome changes in response to inflammatory signals and the surrounding microenvironment. Measurements obtained under standardized culture conditions may therefore not predict cell activity after administration to diseased lungs[
35,
36].
Macrophage-based therapies exploit both physical clearance and local immune regulation. Alveolar macrophages ingest microorganisms, apoptotic cells, cellular debris, and excess surfactant. They then couple lysosomal degradation to cytokine production, lipid metabolism, and inflammatory resolution[
30,
31].
In
Csf2ra-deficient mice, genetically corrected macrophages engrafted within the alveolar compartment and restored granulocyte–macrophage colony-stimulating factor signaling. The transplanted cells maintained surfactant clearance and produced durable improvement in pulmonary alveolar proteinosis[
37].
Regenerative cell therapies aim to reconstruct damaged epithelial lineages. Airway basal cells can generate ciliated and secretory progeny, whereas alveolar type II progenitors produce surfactant and differentiate into gas-exchanging type I cells[
29,
36,
38]. Transplantation of primary or pluripotent-stem-cell-derived basal cells produced durable airway reconstitution in mice. Stem-cell-derived alveolar cells also demonstrated functional engraftment in immunocompetent recipients[
36,
38].
Therapeutic value of context-responsive cellular functions
The principal therapeutic value of mammalian cells lies in their ability to integrate multiple signals from an injured pulmonary microenvironment. Unlike a fixed-dose molecular formulation, a viable cell can alter its activity as local inflammatory, metabolic and tissue-derived signals change. The resulting response, however, depends on the identity and functional state of the administered cell.
MSCs can adjust their paracrine output in response to tissue injury, potentially coordinating epithelial protection, immune regulation, and repair[
35,
39,
40]. Macrophages combine physical clearance with antigen processing, lipid metabolism, and immune signaling[
30,
31]. Epithelial progenitors provide a different benefit by engrafting and replacing damaged functional cells[
29,
36,
38], which has potential usage in complex lung diseases that cannot be addressed through a single molecular target.
Donor variability, phenotypic drift, and engraftment safety
Mammalian cell products are difficult to standardize because their activity varies with donor, tissue source, passage number, expansion conditions, and cryopreservation[
30,
35,
39]. These variables can alter cell identity and potency without producing an obvious change in viable-cell number. Surface-marker profiles and viability measurements are therefore insufficient unless they are linked to the intended therapeutic function.
The diseased lung can further modify cell behavior after administration. Inflammatory signals may suppress MSC-mediated repair, alter the composition of the MSC secretome, or shift macrophages toward tissue-damaging or pro-fibrotic states. Epithelial progenitors may fail to engraft, differentiate incorrectly, or lose regenerative capacity. Context responsiveness is therefore both a therapeutic strength and a source of functional unpredictability.
Physical and biological risks also differ among cell classes. Cell aggregation and loss of viability can produce uneven pulmonary distribution and may obstruct narrow airways. For proliferative or pluripotent-stem-cell-derived products, ectopic tissue formation, inappropriate differentiation, and tumorigenicity require particular attention.
Conversely, rapid clearance creates the opposite problem by removing therapeutic cells before they produce sufficient activity. Product development must therefore balance functional persistence against the risks associated with prolonged residence or uncontrolled engraftment. Release criteria should combine cell identity and viable recovery with mechanism-specific potency assays, such as paracrine activity for MSCs, phagocytic, or surfactant-clearance capacity for macrophages, and lineage-specific engraftment potential for epithelial progenitors.
Acellular biogenic systems and derivatives
Acellular biogenic systems include mammalian EVs, bacterial membrane vesicles, cell membrane-derived vesicles, therapeutic viruses, and bacteriophages. Although these systems lack autonomous metabolism, their functions depend on source-derived membranes, molecular cargo, and host-recognition machinery. Mammalian and bacterial vesicles mainly transport protected cargo, act as receptor decoys or stimulate immune responses. By contrast, viruses and bacteriophages use host recognition and cellular entry to deliver genes, support biological amplification, or kill pathogens. These platforms avoid some risks associated with viable-cell therapies but present challenges in controlling heterogeneity, defining cargo, managing immunogenicity and infectivity, and maintaining functional stability. Figure 3 summarizes these mechanisms and their principal limitations across platforms.
Mammalian EVs
Functional basis and classification of mammalian EV therapeutics
Mammalian EVs are naturally released, non-replicating particles enclosed by a lipid bilayer. They contain proteins, lipids, nucleic acids, and metabolites whose composition reflects the identity and physiological state of the producer cell[
10,
41–
43]. Consistent with MISEV2023, EV is used as the general term, whereas designations such as exosome require evidence of the corresponding biogenesis pathway[
10]. After pulmonary deposition, EVs may bind surface receptors, undergo endocytosis, or fuse with recipient-cell membranes.
Endogenous regulation and defined-cargo delivery
Source-cell-derived regulatory EVs transfer composite endogenous cargo that can modify inflammation, oxidative stress, cellular metabolism, and tissue repair. In lipopolysaccharide-challenged mice, inhaled MSC-derived EVs reduced inflammatory-cytokine production and pulmonary tissue injury. These effects were associated with a shift in alveolar macrophages toward antioxidant and reparative states[
44]. In a murine model of allergic asthma, intranasally administered MSC-derived EVs expanded interleukin-10-producing interstitial macrophages and reduced airway inflammation, mucus hypersecretion, and hyperresponsiveness[
45].
Engineered EVs instead use a predefined cargo to produce a more specific therapeutic mechanism. MSC-derived EVs enriched with miR-146a-5p suppressed interleukin‑1 receptor‑associated kinase 1 (IRAK1)–tumor necrosis factor receptor‑associated factor 6 (TRAF6) signaling, nuclear factor kappa‑light‑chain‑enhancer of activated B cells (NF-κB) activation, and NLR family pyrin domain‑containing protein 3 (NLRP3) inflammasome activity in a mouse model of allergic airway inflammation[
46]. A more complex example used macrophage-derived EVs to combine C-X-C motif chemokine ligand 12 (CXCL12) interception with delivery of the programmed death-ligand 1 (PD-L1) inhibitor BMS-202. Following nebulized administration in a lung-metastasis model, the treatment altered local macrophage phenotypes and disrupted the pulmonary pre-metastatic niche[
47].
Membrane-protected multimolecular delivery
EVs protect multimolecular cargo and retain selected interactions associated with the source cell. They can deliver proteins and RNAs that are unstable in extracellular fluid or inefficiently internalized as free molecules, without requiring survival or proliferation after deposition[
43,
48]. Their ability to carry multiple molecular species can support coordinated regulation of several disease pathways. EVs may also retain source-cell-associated surface molecules that promote interactions with pulmonary epithelial, endothelial or immune cells[
41,
42,
49]. Unlike living cells, EVs cannot adjust transcription or replenish cargo after administration; their activity declines as the vesicles are cleared or degraded.
Product variability and barriers to reproducible EV function
EV heterogeneity begins with the producer cell and is further shaped by culture, isolation, and purification[
10,
11,
50]. Particle count and total protein are incomplete measures of activity because functional molecules may be concentrated in minor vesicle populations, while lipoproteins, protein aggregates and free nucleic acids can confound analytical assays. Cellular uptake also does not ensure cytosolic delivery, as endosomal retention and lysosomal degradation can inactivate cargo. Source-cell selection must therefore account for unwanted pro-inflammatory, procoagulant, fibrotic or oncogenic components in the final product.
Bacterial membrane vesicles
Functional basis and classification of bacterial membrane-vesicle therapeutics
Bacterial membrane vesicles are non-replicating lipid-bilayer particles released by Gram-negative and Gram-positive bacteria[
51,
52]. Gram-negative species produce OMVs, whereas Gram-positive bacteria release vesicles across the cytoplasmic membrane and cell wall[
51,
52]. These vesicles carry bacterial phospholipids, lipopolysaccharide or lipoteichoic acid, peptidoglycan fragments, proteins, antigens, and nucleic acids[
53]. These surface patterns allow the vesicles to deliver cargo while simultaneously stimulating innate immune responses[
53]. Unlike live bacteria, these non-replicating particles avoid colonization risks.
Innate immune priming and antigen-presenting mechanisms
Innate immune-activating vesicles act mainly through pattern-recognition pathways rather than pathogen-specific antigens. Intranasal
Escherichia coli (E. coli) OMVs protected mice against multiple influenza strains by recruiting pulmonary macrophages and inducing type I interferon responses[
54]. The antigen-independent effect was associated with macrophage priming and neutrophil recruitment in the lower respiratory tract. Translation will require careful control of exposure because the same TLR-dependent pathways can increase cytokine release, vascular permeability, and lung injury.
Antigen-presenting and therapeutic-payload vesicles pair the bacterial membrane’s natural adjuvant activity with a specific target payload. In a mouse model of metastatic lung cancer, intranasally administered glycine-induced
E. coli OMVs activated alveolar macrophages and enhanced tumor-antigen presentation, acting as an
in situ vaccine[
55]. Other approaches display defined antigens in repetitive surface arrays to elicit humoral and cellular responses[
56,
57]. Here, the vesicle membrane acts as a bio-adjuvant that facilitates endocytosis by dendritic cells and macrophages, while the surface-exposed or encapsulated cargo drives antigen-specific T-cell responses against the target pathogen or tumor.
Coupled antigen delivery and mucosal adjuvanticity
Bacterial vesicles combine multivalent antigen display with innate immune stimulation in a non-replicating particle. Membrane-bound antigens can retain native conformations, while bacterial pathogen-associated molecular patterns activate local antigen-presenting cells and mucosal immunity[
51,
53,
56,
57]. The absence of reproductive autonomy removes the risks of airway colonization and phenotypic drift associated with live bacterial products. The same intrinsic immunogenicity that supports mucosal vaccination and tumor immunotherapy, however, also narrows the margin between an effective and an injurious pulmonary dose.
Constraints on inflammatory safety and compositional consistency
The microbial components that provide therapeutic benefits from bacterial vesicles can also lead to pulmonary toxicity. Key elements such as lipopolysaccharides, lipoproteins, peptidoglycan fragments, and bacterial nucleic acids are critical in enhancing antigen presentation and activating the innate immune system. However, prolonged exposure to these components can lead to increased cytokine release, higher vascular permeability, and possible tissue damage. Additionally, preparations might contain toxins, virulence factors, antimicrobial resistance genes, or leftover nucleic acids[
51–
53]. Since the composition of vesicles can differ based on strain, growth phase, and culture conditions, the concentration of particles alone is an unreliable indicator of potency[
51,
57]. Genetic attenuation or chemical detoxification can reduce inflammatory activity, although these modifications may also weaken adjuvant function.
Cell membrane-derived vesicles
Functional basis and classification of reconstructed membrane vesicles
Cell membrane-derived vesicles are acellular structures assembled from isolated plasma membranes[
10,
58]. Their activity depends on the retention of source-cell lipids, glycans, adhesion molecules, and surface receptors, while some formulations also encapsulate exogenous cargo. These products are classified here as receptor-decoy systems that intercept pathological ligands or cell-mimetic systems that deliver a defined therapeutic cargo.
Receptor-decoy and cell-mimetic cargo-delivery mechanisms
Decoy systems use retained membrane receptors to intercept a pathological ligand before it reaches viable target cells. Nanodecoys derived from human lung spheroid cells displayed angiotensin-converting enzyme 2 and bound severe acute respiratory syndrome coronavirus 2. In mice, inhaled delivery accelerated clearance of virus-mimicking particles; in infected cynomolgus macaques, it reduced viral burden and lung injury[
59]. The multivalent membrane presents several copies of a physiological receptor in its native lipid environment, enabling competitive sequestration without introducing living target cells. Decoy performance depends on receptor density, affinity, accessible orientation, and whether viral evolution or tissue distribution reduces competitive binding.
Cell-mimetic cargo-delivery systems use reconstructed membranes to protect and transport a defined therapeutic. Lung epithelial cell membrane-derived vesicles carrying anti-miR-155 oligonucleotides increased pulmonary delivery and restored suppressor of cytokine signaling 1 in lipopolysaccharide-injured mice. Treatment lowered inflammatory cytokines and reduced tissue injury compared with free or alternative formulations[
60]. Here, anti-miR-155 directly inhibits the pro-inflammatory microRNA to derepress suppressor of cytokine signaling 1 and resolve downstream inflammation, while the membrane shell enables cellular entry.
Receptor display and cargo loading
Reconstructed membranes can display multiple copies of source-cell receptors within a lipid bilayer, enabling multivalent binding to viruses, toxins, or soluble inflammatory mediators[
58,
61]. Selecting a source cell with an appropriate receptor profile can therefore align the vesicle surface with a particular pathological target. Their coreless architecture also supports flexible cargo loading. Hydrophilic agents can be encapsulated within the aqueous lumen, while hydrophobic compounds can be incorporated into the lipid bilayer. This allows receptor-mediated interception or cellular interaction to be combined with delivery of an independently selected therapeutic cargo.
Constraints on membrane orientation, source-cell safety, and stability
Membrane reconstruction can disturb lipid-bilayer organization, affecting protein conformation, glycosylation, and membrane sidedness[
58,
61]. As a result, some receptors may face the vesicle lumen or become inaccessible through steric obstruction, reducing the receptor density available for multivalent binding. Reconstruction methods can also introduce variation in particle size, lamellarity, payload loading, and residual intracellular proteins. The source cell presents a separate safety concern. Conventional particle counts do not detect co-isolated alloantigens, inflammatory mediators, or tumor-associated proteins. Moreover, structural integrity does not necessarily indicate retained biological activity. Lipid oxidation, surface proteolysis, or cargo leakage may impair function before visible particle degradation occurs.
Therapeutic viruses and bacteriophages
Host-dependent entry, replication, and genome delivery
Therapeutic viral platforms depend on host cells for entry, genome delivery and, in some cases, replication. Viruses lack the metabolic machinery required to perform these processes independently. Their activity therefore depends on susceptible bacterial or mammalian hosts[
12,
62]. In respiratory applications, bacteriolytic phages multiply in target bacteria, whereas replication-defective vectors deliver transgenes without producing viral progeny. Replication-competent oncolytic viruses instead amplify preferentially in malignant cells[
63].
Phage lysis, gene transfer, and oncolytic amplification
Bacteriolytic phages recognize receptors on susceptible bacterial strains and inject their genomes into the cells. Newly produced phages are released through host-cell lysis. This cycle can sustain antibacterial activity while susceptible bacteria remain. However, it also makes host range an important determinant of effective dosing. In a study of nine adults with cystic fibrosis and drug-resistant
Pseudomonas aeruginosa, personalized nebulized phages were well tolerated and followed by a median 10
4 colony-forming-unit mL
−1 decline in sputum bacteria. Concomitant care and the uncontrolled design prevent firm efficacy conclusions[
64]. A separate double-blind phase 1b/2a study involving nine adults showed that nebulized BX004-A reached the lower airways and met safety and tolerability endpoints, although efficacy outcomes remained exploratory[
65].
Replication-defective viral vectors deliver genetic payloads without producing infectious progeny. Tissue tropism governs cellular entry and regional lung delivery, whereas promoter activity and genome persistence determine expression duration[
12]. For example, aerosolized 4D-710 utilizes an engineered adeno-associated virus (AAV) capsid to deliver a truncated cystic fibrosis transmembrane conductance regulator cassette, demonstrating functional ion channel rescue in human airway cultures and dose-dependent expression in non-human primates[
66]. Similarly, a vascular-tropic AAV carrying fibroblast growth factor 12 reached pulmonary arterial smooth muscle cells to prevent and reverse pulmonary arterial hypertension in rodent models[
67]. Also, preclinical studies of third-generation rSIV.F/HN lentiviral vectors carrying full-length cystic fibrosis transmembrane conductance regulator (CFTR) have demonstrated durable airway expression and potential for repeat dosing and have progressed toward first-in-human evaluation for cystic fibrosis[
68].
Replication-competent oncolytic viruses replicate within malignant cells, causing tumor debulking, and promoting antitumor immunity. Tumor-cell lysis releases antigens, viral progeny and danger signals that recruit innate and adaptive immune cells[
63]. In an immunocompetent small-cell lung cancer model, intrapulmonary myxoma virus preferentially replicated in tumor tissue, induced necrosis and immune-cell infiltration, and prolonged survival; activity was also reported in patient-derived models[
69].
Target recognition and biological amplification
The specific site where viral systems deliver their genomes or initiate replication is fundamentally dictated by target recognition. Bacteriophages, for example, bind to specific bacterial receptors to combat drug-resistant pathogens. However, the overall therapeutic response is heavily influenced by strict host specificity and phage resistance[
64,
70]. In a different approach, replication-defective vectors offer a distinct advantage by sustaining therapeutic protein expression long after the initial airway particles have been cleared[
12,
66,
67]. Oncolytic viruses introduce yet another layer of therapeutic amplification, actively coupling direct tumor-cell lysis with local antigen release and robust immune activation[
63,
69]. While all these viral mechanisms rely on receptor-dependent entry to breach the target cell, they diverge significantly regarding their specific cellular targets, duration of action, and the downstream consequences of replication.
Constraints on host range, repeat dosing, and replicative safety
While bacteriophages offer highly targeted therapy, their clinical efficacy is inherently bottlenecked by an extremely narrow host specificity and the rapid emergence of receptor-mediated resistance, particularly within complex, heterogeneous infections[
62,
70,
71]. Inducing rapid bacterial lysis also carries the secondary risk of flooding the local tissue environment with endotoxins and cellular debris. Personalized phage cocktails present practical challenges for clinical translation. Their variable composition complicates large-scale manufacturing and makes batch-to-batch standardization difficult. Mammalian viral vectors have different limitations. Replication-defective vectors have restricted packaging capacity, and their tissue tropism can be difficult to predict. Neutralizing antibodies may also reduce the effectiveness of repeat dosing[
64,
69]. Integrating vectors carry a risk of insertional genotoxicity. AAV preparations may contain empty capsids, or truncated genomic payloads, reducing product consistency[
72]. Finally, oncolytic viruses require a balance between immune control and sustained viral activity. Premature clearance by the host immune system severely blunts therapeutic activity. Conversely, allowing viral replication to proceed unchecked can trigger widespread inflammation, collateral off-target tissue damage, and hazardous viral shedding[
63].
Translational limitations shared across all classes
Nominal input does not adequately describe the activity of biogenic therapeutics. Product specifications need to identify the fraction that remains functional and reaches the relevant respiratory compartment[
1,
6]. The active unit differs by mechanism: viable cells must retain effector secretion, motility, phagocytosis, or regenerative capacity; viral systems must retain infectivity or transduction; and membrane vesicles must preserve accessible receptors and functional cargo release[
10–
12]. Physical counts remain useful for identity and recovery, but they cannot substitute for a mechanism-linked potency assay.
Biological function can decline at several points between manufacture and tissue exposure. Donor or batch heterogeneity, storage and formulation stress, mucociliary clearance, immune neutralization, and disease-related airway obstruction all reduce the active fraction of the administered dose[
1,
9–
12]. The relevant failure mode is platform specific. For example, a membrane vesicle may retain its particle count after a conformational change has abolished receptor binding. Translation therefore requires potency measures tied to the intended lung compartment and evidence that the responsible function survives formulation, aerosolization, and deposition.
Engineering strategies for inhalable living and bio-derived systems
Living organisms intrinsically operate through adaptive sensing, immunomodulation, motility, and site-specific payload secretion[
25,
73]. Acellular bio-derived products lack such adaptive self-renewal but successfully retain vital biological recognition and transport capabilities through their source-associated membranes. However, inherent functional heterogeneity complicates both approaches. Bacterial and therapeutic cell efficacies are highly sensitive to donor sources and culture conditions[
16,
74–
77]. Furthermore, physical stressors like nebulization can severely compromise microalgal motility[
21], while phage infectivity and EV composition remain highly dependent on isolation techniques and storage formulations[
65,
78–
80]. Overcoming these translational hurdles in pulmonary medicine demands precise characterization of product identity, ensuring sustained on-target activity without functional loss during aerosolization.
Transforming biological systems into safe and effective therapeutics requires a multi-layered, synergistic engineering strategy. These four engineering modifications are closely interconnected, and none can be optimized in isolation. Specifically, bioprogramming enhances the drug delivery system’s sensing capabilities, effect output, and system closure at a fundamental level; customized carrier and biointerface design helps the drug cross the lung barrier and activate resident immune cells; furthermore, to achieve clinical application, formulation engineering is needed to maintain the efficacy of the therapeutic drug during storage, and precise nebulization engineering ensures device compatibility and lung deposition in the target area (Tables 2 and 3).
Engineering biological functions and safety control
Biological programming modifies a living system, virus or producer cell to establish defined sensing, effector, or safety functions. The appropriate strategy depends on autonomy, replication, genetic accessibility, and the required duration of action. Bacteria, therapeutic viruses and bacteriophages are usually modified at the genome level. Mammalian cells can undergo stable genetic modification or transient phenotypic and metabolic reprogramming. EVs and OMVs are programmed indirectly through changes to producer cells or vesicle-biogenesis pathways before collection.
Bacteria, viruses, and bacteriophages
Bacteria, viruses and bacteriophages are genome-programmable systems, but their engineering priorities differ. Bacteria can autonomously sense environmental signals and synthesize effectors, so chassis selection, attenuation, biocontainment, and programmed secretion are central. Mazzolini et al. selected a genome-reduced, lung-adapted
Mycoplasma pneumoniae chassis, deleted the virulence-associated genes
mpn372 and
mpn133, and programmed the strain to express pyocin L1 and three
Pseudomonas aeruginosa biofilm-degrading enzymes. Following intratracheal administration, the engineered bacteria suppressed pulmonary infection and disrupted endotracheal-tube biofilms[
19]. Rottinghaus et al. constructed a chemically and temperature-responsive CRISPR-Cas9 kill switch in
E. coli Nissle, improving genetic stability and reducing escape under biocontainment conditions[
77]. Beyond applications in infectious diseases, cell reprogramming has also been used to diagnose lung diseases. Researchers combined nanoluciferase release with specific tumor-peptide recognition in
Lactobacillus plantarum, transforming it into an
in vivo biosensor that precisely locates lung tumors in mice without highly invasive manipulation[
81].
Unlike bacterial systems, phage engineering primarily targets two aspects: host recognition and intracellular killing. For example, Gencay et al. screened over 150 wild-type candidate phages to identify those suitable for structural and genomic modification. They then altered their tail fibers and introduced them into a CRISPR-Cas system, successfully expanding the targeting range of
E. coli and inhibiting tolerant subsets[
82]. Although these findings are based on intestinal models rather than lung models, their genetic engineering approaches can be adapted for the engineering of respiratory phages. Furthermore, the design of therapeutic viruses for mammals requires entirely different directions of modification, particularly in the control of replication regulation, tissue specificity, and transgene expression. Researchers introduced a drug-responsive circuit into varicella-zoster virus, enabling the virus to not only maintain its original anti-tumor efficacy but also externally regulate viral replication and therapeutic release[
83].
Cellular functional programming and phenotype engineering
Sophisticated engineering strategies can enhance the secretory characteristics, metabolic tolerance, and immunomodulatory functions of mammalian cells in the treatment of lung diseases. Unengineered MSCs and macrophages, as donor cells, exhibit significant differences and struggle to survive long-term in the complex inflammatory microenvironment of the lungs, greatly diminishing their therapeutic efficacy. To overcome these challenges in translational research, current biological programming strategies focus on achieving precise mechanistic functions. Recent studies have shown that pre-aerosolization modification of MSCs enables intracellular overexpression of angiopoietin-1. Compared to the delivery of wild-type MSCs, engineered cells are able to repair airway tissue more efficiently and suppress allergic reaction chains[
84]. Macrophages engineered to continuously secrete interleukin-4 promoted a reparative phenotype after intratracheal transplantation and reduced acute lung injury[
85], whereas introduction of
Csf2ra restored GM-CSF receptor signaling and alveolar surfactant clearance in a model of pulmonary alveolar proteinosis[
37]. Metabolic programming provides a complementary approach. Increasing glycogen synthesis improved early MSC survival and therapeutic efficacy in bleomycin-induced pulmonary fibrosis[
87].
Genetic programming of microalgae has rarely been evaluated in pulmonary models. Most current studies exploit native motility and add therapeutic cargo by surface conjugation or biohybrid assembly[
25]. It remains unclear whether engineered expression can be stabilized while preserving motility and whether modified algae can be adequately contained after inhalation. These questions need to be resolved before genetic programming can be considered an established strategy for pulmonary microalgal therapy.
Programming of cell-free biological derivatives
Bacterial OMVs and mammalian EVs retain selected membrane components and cargo from their source cells without the proliferation associated with living therapeutics. Once released, they cannot replenish cargo or respond autonomously to disease signals. Programming therefore occurs before collection through modification of producer cells, vesicle biogenesis, or endogenous cargo-sorting pathways. Post-production loading and direct surface modification are treated separately because they alter the released vesicle rather than its formation.
For OMVs, producer-strain engineering must balance vesicle yield, defined antigen or protein incorporation and inflammatory safety. Deletion of
msbB, which encodes a lipid A acyltransferase, generated OMVs containing under-acylated lipid A with reduced TLR4-dependent inflammatory activity while retaining antitumor immune stimulation[
88]. Because this modification changes vesicle composition during formation, it is a producer-level intervention. Modular ligation onto isolated OMVs instead changes the released product directly.
For EVs, the first challenge is to couple increased production with selective enrichment of a predefined cargo. The EXOtic platform combined regulators of EV biogenesis with an L7Ae and C/D-box interaction for mRNA recruitment and connexin-43-mediated functional transfer[
89]. Cellular nanoporation similarly stimulated EV production while enriching transcribed therapeutic mRNAs, producing substantial increases in EV yield and associated mRNA[
90]. These strategies treat EV generation and cargo incorporation as linked producer-cell processes rather than independent downstream operations.
Active sorting systems can replace stochastic incorporation with sequence- or scaffold-guided enrichment. A pre-miR-451-derived stem-loop enabled efficient packaging of therapeutic siRNA into small EVs and reduced the dose required for gene silencing[
91]. Protein cargo can be positioned using EV-enriched scaffolds with defined topologies: PTGFRN supports high-density membrane display, whereas BASP1-derived sequences enable luminal loading of soluble proteins and RNA-binding complexes[
92,
93]. Producer-cell programming thus controls EV formation and endogenous cargo incorporation; pharmacokinetic shielding, target recognition, and fusogenic entry require direct modification of the released vesicle surface.
Engineering cargo and biointerfaces
A programmed carrier does not by itself determine how much exogenous cargo is loaded, where the cargo resides or whether it remains functional after pulmonary delivery. The unmodified carriers are highly prone to being encapsulated by respiratory mucus after administration, or lose their stability due to interaction with the surfactants in the lung fluid, ultimately being rapidly phagocytosed and cleared, and thus unable to reach the intended target sites. Researchers employed two engineering strategies to address these physiological challenges. The first one was to optimize the separation of the internal payload through drug delivery engineering, and enhance the drug delivery kinetics and maintain the long-term drug release characteristics. The second one was to use bio-interface engineering to reshape the outer structure of the carrier by regulating specific surface features such as viral capsids, glycoproteins, or lipid regions, enabling the carrier to avoid immune surveillance during its movement in the pulmonary microenvironment and reducing immune responses. Although the payload and surface characteristics continuously interact, they were strategically adjusted to respectively address distinct delivery failure modes.
Cargo engineering: from stable encapsulation to intracellular delivery
The architecture of cargo integration depends on the physical and chemical properties of the cargo as well as the physiological limitations of the selected carrier. For instance, the main purpose of encapsulating low-molecular-weight drugs is to construct a safe storage depot to reduce off-target toxicity and prevent rapid clearance from the entire body. The delivery of sensitive macromolecules faces completely different challenges: biological agents such as nucleic acids and therapeutic proteins, for example, need to have strong protective mechanisms to resist the widespread nucleases and proteases in the extracellular environment, and at the same time, they need specific mechanisms to ensure their precise targeting in the cytoplasm or nucleus. Furthermore, whenever live cellular or microbial platforms serve as the chassis, loading protocols face a strict boundary condition: the physical encapsulation process must leave the vector’s fundamental survival, functional polarization, and chemotactic mobility entirely intact. Viral vectors are constrained by genome capacity and expression control. For EVs and OMVs, payload location in the lumen or membrane must be related to functional release after pulmonary administration.
Cargo loading in bacterial carriers
Bacterial cargo engineering addresses a problem that chassis programming alone cannot solve: sufficient exogenous drug must be retained and released without impairing bacterial viability, motility, or adhesion. One approach is intracellular compartmentalization of formulated drugs. Paclitaxel-loaded liposomes introduced into bacteria created a paclitaxel-in-liposome-in-bacteria system in which the liposome protected the chemotherapeutic payload and the bacterial carrier increased pulmonary accumulation after inhalation[
94]. Intracellular cargo can be metabolized, expelled or diluted during bacterial division, so cargo per viable CFU and release kinetics need to be related to growth and motility.
Large or chemically fragile cargoes can instead be attached to the bacterial surface so that propulsion and treatment remain functionally separated. Click-mediated attachment of curcumin-loaded hybrid-membrane nanoparticles to magnetotactic bacteria preserved bacterial motion while adding magnetic guidance, inflammatory neutralization, and local drug delivery in pneumonia[
95]. Magnetically steerable bacteria carrying nanoliposomes provide related evidence for externally controlled and stimulus-responsive release[
96]. Protective or bioactive coatings can be used when direct drug exposure threatens the bacterial chassis. A chitosan, hyaluronic acid and ononin coating around
Lactobacillus rhamnosus combined probiotic activity with macrophage-associated targeting and inflammatory regulation[
97]. Coatings also change immune and barrier interactions, making it necessary to distinguish their effects on cargo delivery from their effects on the carrier surface.
Phage selection and combination design
For bacteriophages, the active payload is commonly a selected phage or defined phage mixture rather than an externally loaded drug. The first cargo-level problem is narrow host range. Respiratory infections contain genetically heterogeneous
P. aeruginosa,
Klebsiella pneumoniae, or
Acinetobacter baumannii populations that may not be covered by a single phage. Cocktails and personalized phage sets therefore combine complementary receptor usage and susceptibility profiles. Nebulized BX004-A and personalized inhaled phage regimens provide early clinical evidence in cystic fibrosis[
64,
65,
98], whereas screened combinations against multidrug-resistant
K. pneumoniae support the same design principle in experimental pneumonia[
99,
100]. Aerosolization can change cocktail composition; loaded, emitted, and deposited plaque-forming unit (PFU) therefore need to be quantified for each component.
A second problem is the emergence of phage-resistant populations under lung-specific conditions. Cargo design can coordinate phages with antibiotics so that receptor loss or other resistance mechanisms impose therapeutically useful trade-offs. Lung adaptation studies showed that phage-resistant
P. aeruginosa could regain susceptibility to antibiotics and that sequential phage exposure followed by antibiotic treatment cleared infection
in vivo[
101]. Synergy has also been reported for phage combinations with ceftazidime or tobramycin in cystic-fibrosis-associated models and in precision treatment of refractory
A. baumannii pneumonia[
102,
103]. The relevant design variable is therefore the combination and administration sequence, together with the active dose of each phage.
Phage-derived depolymerases and lysins can degrade bacterial capsules, biofilms, or cell walls and thereby increase access by immune cells and antibiotics[
104–
107]. These purified enzymes are distinct pulmonary biologics. They constitute phage cargo only when the administered intact phage genetically encodes them; otherwise, they should be discussed as phage-derived molecular therapies.
Therapeutic viruses as genetic cargo carriers
Therapeutic viruses primarily deliver genetic cargo, so packaging capacity, expression duration, and disease-site control determine cargo design. In cystic fibrosis, the size of the CFTR coding and regulatory sequence constrains compact AAV genomes. The aerosolized candidate 4D-710 combines an evolved AAV capsid with a partially shortened CFTR transgene that fits within the vector while retaining functional activity in airway epithelial models and producing dose-dependent expression after aerosol administration in nonhuman primates[
65]. Lentiviral vectors can accommodate larger or full-length cassettes and provide longer expression, but integration risk, airway stem-cell function and redosing must be assessed[
108].
When mucus or airway fluids prevent a viral vector from reaching cells in a transcriptionally active form, cargo protection can be coupled to a biological carrier. EV-associated AAV6 increased reporter expression in mucus-covered airway cultures and mouse airways compared with standard AAV6[
109,
110]. Motile cyanobacteria conjugated to PEGylated oncolytic adenoviruses similarly transported viral cargo through pulmonary barriers and enabled local release in experimental lung tumors[
111]. Evaluation of these hybrid systems requires evidence that both carrier function and viral infectivity survive assembly and pulmonary delivery.
Oncolytic viruses can also be armed with cytokines, suicide genes, or checkpoint-modulating agents when direct lysis is insufficient in an immunosuppressive lung tumor. Adenoviral thymidine-kinase delivery followed by valacyclovir induced immunogenic tumor-cell death in patients with non-small-cell lung cancer[
112]. Adenoviruses and vaccinia viruses carrying interleukin‑12 (IL-12), tumor necrosis factor‑α (TNF-α), interleukin‑2 (IL-2), or local PD-L1 blockade have further increased antitumor immunity in lung-cancer models[
113–
115]. Arming can extend a cytolytic vector into a source of local immunomodulation. Evidence from non-inhaled administration, however, does not establish aerosol compatibility or pulmonary distribution.
Cargo-loaded therapeutic cells
Mammalian cells can transport drugs or nanoparticles to inflamed, fibrotic or metastatic lungs, but loading may impair viability, migration, adhesion, phagocytosis, or immune phenotype. Cargo localization should therefore match the intended release pathway. M1 macrophages carrying liposomal resiquimod (R848) intracellularly and a protease-sensitive ravtansine (DM4) prodrug on the membrane used spatial compartmentalization to combine immune stimulation with lesion-responsive cytotoxic release in lung metastasis[
116]. Monocyte-derived multipotent cells bearing astaxanthin- and trametinib-loaded particles similarly homed to injured lungs and released cargo in response to matrix metalloproteinase‑2 (MMP-2) in pulmonary fibrosis[
117].
Surface backpacks offer a partial membrane patch when internalization would destroy the cargo or interfere with phagocytic function. Interferon‑gamma (IFN-γ)-loaded polymeric backpacks resisted uptake by their carrier macrophages and provided sustained phenotypic stimulation[
118], whereas magnetic backpacks added spatial localization and imaging[
119]. Most studies used systemic cell administration rather than inhalation and therefore establish design principles for pulmonary lesions, not aerosol compatibility. Directly inhaled cargo-bearing cells have rarely been tested; essential post-loading endpoints include viable-cell recovery, aggregation, phenotype, migration, and cargo release.
Microalgal cargo transport
Microalgal cargo design balances payload amount against preservation of motility. Intracellular drug loading can expose algae to toxic compounds, while extensive surface coverage may damage flagella, increase hydrodynamic drag, or promote macrophage uptake. Most pulmonary constructs therefore attach protected therapeutic modules to part of the algal surface, allowing propulsion and cargo release to be optimized separately.
Neutrophil-membrane-coated antibiotic nanoparticles attached to motile microalgae retained movement, distributed through lung tissue and improved treatment of acute
P. aeruginosa pneumonia after intratracheal administration[
24]. Red-blood-cell-membrane-coated doxorubicin nanoparticles used the same modular principle to suppress lung metastasis. A subsequent inhalable design attached platelet-membrane-coated vancomycin particles to picoeukaryotic algae and preserved motility and cargo function after nebulization[
25]. Surface loading altered both transport and therapeutic function, so cargo per viable alga must be interpreted together with swimming speed, particle retention, and post-release activity.
Microalgae can also transport biological cargo. Motile
Synechococcus conjugated to tumor-responsive oncolytic adenoviruses improved viral distribution and local release in lung-cancer models[
111,
120]. Photosynthetic oxygen generation represents a separate mechanism and currently depends on auxiliary illumination. The more direct pulmonary evidence for cargo transport involves antibiotic, chemotherapeutic and viral payloads.
Cargo loading in EVs and OMVs
EVs and OMVs are non-replicative membrane carriers, but their native cargoes are heterogeneous and often insufficient for a defined therapeutic purpose. For mammalian EVs, fragile RNA cargo requires efficient loading, protection during aerosolization, and functional release into the correct lung cell. Lung-derived EVs delivered functional mRNA and protein to bronchioles and parenchyma after nebulization[
121], whereas vibrating-mesh nebulization of siMyd88-loaded small EVs reduced lipopolysaccharide (LPS)-induced acute lung injury[
49]. IL-12 mRNA-loaded EVs enabled local cytokine production in lung tumors with lower systemic exposure[
122]. Functional assessment requires loading efficiency and vesicle recovery to be linked to RNA integrity and expression or silencing in target cells.
Cargo loading alone does not ensure delivery to the intended pulmonary population. Inhalable EVs carrying mesothelin-specific CAR mRNA and displaying anti-CD206 scFv generated CAR macrophages
in situ[
123,
124], while helper-lipid modification redirected tumor-derived EVs and their siPD-L1 cargo toward lung metastases[
125]. In these coupled designs, RNA provides the therapeutic function and the displayed ligand, or helper lipid determines access to the recipient-cell population.
OMV cargo engineering must balance defined payload incorporation with control of endogenous bacterial components. Engineered
E. coli expressing ClyA-OVA257-264 generated inhalable membrane vesicles that co-delivered tumor antigen and innate immune stimuli to pulmonary lymphoid and metastatic sites[
126]. Glycine-induced OMVs altered yield, LPS content and outer-membrane-protein composition, providing a preparation-based route to tune endogenous cargo[
54]. Doxorubicin-loaded OMVs and newer nucleic-acid or immunomodulatory platforms broaden the potential cargo space[
127–
129]. Studies without inhaled administration support the loading method but do not establish pulmonary delivery.
Biointerface engineering
Biointerface engineering changes structures exposed to the surrounding biological environment, including membranes, glycocalyx, capsids, and phage tail fibers. These modifications are used to alter transport through lung barriers, immune recognition, disease-site binding, or cellular entry. Changes confined to the internal payload are considered cargo engineering rather than interface engineering.
Pulmonary barriers and clearance
After deposition, inhaled systems encounter mucus, pulmonary surfactant, and alveolar macrophages before reaching many target cells. Mucus can immobilize viruses, phages, EVs, and microbial carriers; surfactant adsorption can alter membrane interactions; and macrophages may remove foreign systems before they act. Studies using zwitterionic nanoparticle surfaces further show that local pH can alter particle–mucin interactions and mucus penetration[
86]. Interface design at this stage concerns transport and clearance after deposition. Damage generated by the nebulizer or inhaler is addressed separately in Section Aerosolization and pulmonary delivery.
For EVs and viral vectors, the relevant interface is the vesicle membrane or capsid. Donor-cell selection can provide an intrinsically lung-compatible EV surface, as lung-derived EVs showed greater bronchiolar and parenchymal distribution than HEK293-derived EVs or liposomes after nebulization[
121]. Vesicle association improved AAV6 transduction across mucus-covered airway cultures[
109], whereas directed capsid evolution generated an AAV interface with enhanced airway epithelial transduction[
66]. Their performance is better defined by mucus transport, epithelial access, and functional expression than by deposition alone.
Motile bacteria and microalgae can actively redistribute after deposition, but surface loading or coating may increase drag, aggregation, and macrophage capture. Magnetically actuated bacteria carrying hybrid-membrane cargoes and membrane-camouflaged microalgal microrobots illustrate how surface composition can be tuned while retaining motion in lung-relevant fluids[
24,
95]. Surface modification must preserve carrier motility while limiting nonspecific adhesion and phagocytosis. For phages, mucus, hypoxia, and polyamines can alter adsorption to pulmonary pathogens; capsid stability and receptor binding therefore need testing in lung-like media rather than simple buffer[
101].
During the process of directly delivering mammalian cells to the lungs for the treatment of lung diseases, it is impossible to directly modify their extracellular structures because these living carriers cross the respiratory barrier through two natural characteristics: their inherent physical flexibility and specific extracellular matrix. Extensive surface modification would mean interfering with their original chemotactic pathways. To avoid disrupting their tissue integration and directed cell migration, researchers can only prioritize sparse functionalization rather than covering up their structure. On the other hand, OMVs face another predicament: their naturally abundant lipopolysaccharides and surface glycoproteins not only can facilitate the rapid uptake by mucosal immune cells, but can also trigger intense inflammatory clearance responses. Therefore, precise separation techniques are required to handle OMVs, removing specific endotoxins while ensuring that the vesicles can still be recognized by resident macrophages[
55].
Immune recognition and inflammatory control
Successfully crossing the pulmonary immune barrier depends on a deliberate choice between immune evasion and active immune stimulation. Therapeutic cells and gene vectors generally require extensive surface camouflage to survive non-specific phagocytic clearance. Vaccine vectors and bacterial OMVs, by contrast, rely on provoking a robust local inflammatory response to achieve their intended effect. These opposing demands call for precise immune interface engineering, in which the outward-facing molecular features of a vector are carefully adjusted to reshape how host cells interact with it and how strongly inflammation is triggered. At the same time, this immune modulation must leave the platform’s underlying genetic or therapeutic architecture intact.
For bacterial carriers, localized coatings can shield strongly reactive surfaces or neutralize inflammatory mediators while preserving bacterial viability and motion. Hybrid-membrane cargoes attached to magnetotactic bacteria and polysaccharide-based coatings around probiotics illustrate this approach in pneumonia models[
95,
97]. For OMVs, the immune interface is dominated by lipid A, LPS, outer-membrane proteins, and glycans. Glycine-induced vesiculation and endotoxin-free OMV platforms reduce inflammatory burden while preserving antigen presentation or immunotherapeutic activity[
55,
130]. Lipid A or LPS activity, cytokine induction, antigen-presenting-cell activation, and lung histopathology provide complementary measures of this balance.
The desired viral immune interface depends on therapeutic purpose. AAV capsids need to support epithelial entry despite neutralizing antibodies. Oncolytic and vaccine viruses instead rely on controlled local immune engagement. The evolved capsid used in 4D-710 retained airway transduction in the presence of pre-existing neutralizing antibodies[
66]. Cytokine or antigen payloads alter cargo, not the immune interface, unless the capsid or envelope is also modified. Currently, deliberate modifications to the bacteriophage immune interface for respiratory applications are rarely explored. While clinical evaluations rigorously monitor safety parameters such as formulation diversity, post-lysis endotoxin spikes, and collateral shifts in the local microbiome, these factors primarily represent downstream biological consequences. True immune-interface engineering demands a more proactive approach, specifically requiring targeted structural alterations to the viral capsid or exterior surface to directly dictate host interactions.
Instead of fully encapsulating living vectors, strategically decorating a fraction of their exterior allows researchers to embed new immunoregulatory traits without sacrificing endogenous functions. A prime example involves affixing localized therapeutic depots on macrophages. This backpack strategy actively shields the payload but leaves the vast majority of the cellular membrane fully exposed, ensuring unimpeded spatial navigation, and microenvironmental sensing[
118]. Parallel logic is also applicable to the microalgae delivery system. The researchers directly anchored the payload to the surface of the algae. These payloads were disguised on the membranes of neutrophils or platelets. This approach reduced nonspecific phagocytosis while preserving algal propulsion[
24,
25]. When this technology is applied to EVs, the specifically designed surface ligands can cause a change in cell specificity. For example, after the single-chain fragment that binds to the CD206 receptor on the surface is formed, the aerosolized vesicles break away from the conventional clearance pathway and are directed towards the pulmonary macrophage population[
123]. So, evaluating the effects of these interface modifications requires an overall perspective. We need to weigh the engineered immune evasion and regulated cytokine response. At the same time, we also need to consider the critical balance with the basic biological activity of the carrier.
Target recognition and cellular entry
Therapeutic vectors first successfully overcome the mucosal barrier, and then need to have high affinity to specifically bind to the pathogen targets or designated host cells. Interface engineering enhances binding specificity by modifying the outward fusion domain. This engineering ensures that the initial physical contact is transformed into effective intracellular payload release. The biomechanical mechanism that determines this crucial step varies depending on the biologically active system. Bacteriophages rely on highly specific interactions with bacterial surface antigens, while therapeutic viral vectors require a complex receptor-mediated endocytosis process to infect cells. EVs must directly fuse with the host plasma membrane or undergo strong membrane fusion deep within the endosomal cavity to successfully release the payload into the cytoplasm.
For bacteriophages, tail fibers and receptor-binding proteins are the principal target-facing elements. Engineering complementary tail fibers can broaden bacterial coverage and reduce escape, as demonstrated by CRISPR-armed phages selected for distinct
E. coli receptors[
82]. The strategy broadens receptor coverage, but it has not yet been validated directly in pulmonary infection. Viral vectors use an analogous principle at the capsid or envelope. Directed capsid evolution can improve airway epithelial tropism[
66], whereas vesicle association can preserve receptor engagement and functional transduction across airway barriers[
109].
EV target interfaces can be engineered by displaying antibodies, receptor-binding peptides or decoy proteins. Anti-CD206 scFv directed EV uptake toward a defined macrophage population in the lung[
123]. Angiotensin-converting enzyme 2 (ACE2) variants displayed on truncated cluster of differentiation 9 (CD9) scaffolds converted EVs into vesicular decoys for severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2)[
131], while receptor-binding domain (RBD)-tagged EVs combined ACE2-associated targeting with antiviral siRNA delivery[
132]. Multifunctional EVs can combine a targeting domain, cargo-loading scaffold and fusogenic protein to coordinate binding and cytosolic entry[
133]. Evidence generated outside the lung supports the design method but not pulmonary performance.
OMVs can display defined antigens or binding proteins through genetically encoded outer-membrane scaffolds. ClyA-linked SpyTag-SpyCatcher and SnoopTag-SnoopCatcher systems enabled modular attachment of one or more antigens to the OMV surface[
134], whereas inhalable ClyA-antigen membrane vesicles promoted antigen presentation in pulmonary lymphoid tissues[
126]. Direct receptor engineering of inhaled mammalian cells, bacteria and microalgae remains less developed than cargo attachment or native tropism. In living carriers, increasing surface coverage is useful only if essential adhesins, chemotactic receptors, flagella, and glycocalyx function remain intact.
Engineering formulation and functional stability
Formulation requirements differ because the therapeutically active unit differs across platforms. Living products must retain viability and the functions coupled to it; phages and therapeutic viruses must retain infectivity; and membrane vesicles must preserve membrane structure, cargo, and surface markers. Composite constructs add further failure points because the carrier, coating, linker, and payload may deteriorate at different rates. Stability studies should therefore begin with platform-specific failure modes and use assays that measure the function responsible for therapeutic activity. Platform-specific formulation strategies are summarized in Tables 2 and 3.
Living systems: cells, bacteria, and microalgae
Preservation of a living product requires recovery of both viable cells and the functions that determine potency. Relevant functions include membrane integrity, phenotype, migration, and secretion for mammalian cells; metabolic activity, motility, genetic stability, and cargo retention for bacteria; and swimming ability, flagellar integrity, and attached-cargo stability for microalgae. A viable-cell count alone cannot establish functional recovery after storage.
Mammalian cell carriers are usually preserved either as short-term hypothermic suspensions or as long-term cryopreserved products. Short-term storage is commonly performed at 2–8°C in isotonic, clinically compatible preservation solutions, but simple saline or culture medium may not be sufficient because hypothermia can induce ionic imbalance, cell swelling, loss of recovery, and altered paracrine activity. Trehalose-based or commercial hypothermic solutions have been shown to preserve human MSC viability, identity, differentiation capacity, and immunomodulatory properties for up to 72 h at 4°C, whereas post-thaw handling also requires protein-containing or otherwise optimized reconstitution media to avoid rapid cell loss after dilution[
135,
136]. For long-term storage, mammalian cells are typically cryopreserved with controlled-rate cooling, commonly around 1 °C/min, in cryoprotectant-containing media such as 5%–10% dimethyl sulfoxide (DMSO), followed by storage in liquid or vapor-phase nitrogen. DMSO-free formulations containing non-permeating and permeating protectants, such as sucrose, glycerol and amino acids, are also being explored to reduce DMSO-associated toxicity while maintaining acceptable MSC recovery and immunophenotype[
137,
138]. For pulmonary cell carriers, viable-cell recovery needs to be related to migration, adhesion, immune phenotype, secretion, and retention of surface-attached cargo.
Bacterial carriers are generally more tolerant than mammalian cells, but engineered therapeutic bacteria still require preservation of viability and programmed function. For short-term use, bacterial preparations can be stored as refrigerated liquid suspensions, whereas long-term laboratory preservation commonly uses glycerol stocks at −80°C. For deployable biological therapeutics, however, frozen suspensions are less convenient than dried products, so lyophilization, spray drying, fluidized-bed drying, or matrix encapsulation are often used to generate more stable formulations. During drying, cryo- or lyoprotectants such as trehalose, sucrose, maltodextrin, skim milk, milk proteins, albumin, or polysaccharides reduce membrane damage and improve post-rehydration survival. For example, freeze-dried
Lactiplantibacillus plantarum showed improved survival with saccharide protectants, with skim milk, maltodextrin, and sucrose improving post-drying or refrigerated storage viability[
139]. Fluidized-bed dried
Lactobacillus paracasei embedded in dairy matrices retained better viability at 4°C, 25°C, and 37°C when matrix composition and water activity were optimized[
140]. For engineered living materials, trehalose treatment before lyophilization and desiccant storage enabled long-term room-temperature stability, illustrating how intracellular and matrix-level protection can preserve living biocatalytic systems[
141]. For pulmonary bacterial therapeutics, CFU recovery must be linked to growth, motility, genetic stability, and retention of the engineered therapeutic function after reconstitution.
Microalgae require more strain-specific preservation because freeze-thaw tolerance varies markedly among algal species and even among strains. Short-term storage often relies on maintaining cultures at low temperature under appropriate medium, light and nutrient conditions, whereas long-term archiving generally uses cryopreservation in liquid nitrogen or, for selected Chlamydomonas strains, −80°C algae cryopreservation protocols. Boswell et al. showed that −80°C cryopreservation of
Chlamydomonas reinhardtii could recover some clonal populations, but recovery was strain-dependent and genetically mixed populations could be biased after freezing[
142]. For cyanobacteria and eukaryotic microalgae, exopolysaccharide-based cryoprotection improved growth recovery after cryopreservation compared with 5% DMSO for several tested strains, whereas 10% glycerol showed poor recovery in that study[
143]. For pulmonary microrobots, survival alone is insufficient: stored algae must retain swimming ability, flagellar integrity, and attached therapeutic cargo. The marked strain dependence of recovery means that culture-collection protocols cannot be transferred to an engineered pulmonary construct without functional testing.
Infectious systems: phages and therapeutic viruses
Bacteriophages are often more physically robust than enveloped mammalian viruses, but their stability remains highly phage- and formulation-dependent. Liquid phage preparations are commonly stored at 4°C for short-term use, whereas frozen storage at −80°C or dry-state preservation is used for longer-term storage. However, liquid storage alone may be insufficient for clinical distribution, and repeated freeze-thaw cycles or uncontrolled freezing can reduce viable PFU. Dry formulations have therefore been widely explored. In lyophilized phage preparations, disaccharides such as sucrose and trehalose are frequently used as lyoprotectants because they form glassy matrices that reduce capsid and tail protein damage during freezing and drying. Zheng et al. systematically compared saccharides and polyols for bacteriophage freeze-drying and found that vitrified formulations, especially sucrose and trehalose-containing systems, better preserved phage titers during drying and storage; devitrification during storage at higher temperatures was identified as a major cause of phage inactivation[
144]. Earlier studies likewise found that sucrose, gelatin or their combination maintained lyophilized phage viability for months at 4°C, with greater loss at 37°C[
145].
For pulmonary phage products, dry-powder preservation is attractive because it may reduce cold-chain dependence while maintaining a high delivered PFU after reconstitution or dispersion. Spray-dried Pseudomonas phage powders containing trehalose and leucine maintained phage viability for up to 12 months under vacuum at 4°C or 20°C, although stability depended on the phage type and leucine content[
146]. Anti-Acinetobacter phage powders formulated with trehalose, mannitol and L-leucine further showed that excipient ratios, humidity and storage temperature strongly influence both titer stability and powder handling[
147]. Thin-film freeze-drying has also been used to manufacture phage powders. A sucrose-leucine matrix with an appropriate buffer system protected T7 phage during freezing and sublimation[
148], and a more recent mycobacteriophage D29 dry-powder study showed that trehalose and polyvinylpyrrolidone (PVP)-containing amorphous matrices preserved phage activity for months at 4°C and 22°C under low humidity, whereas 40°C storage caused rapid loss of activity[
149]. PFU recovery therefore needs to be considered alongside adsorption, host range and stability under the intended humidity and temperature conditions.
Therapeutic viruses require more platform-specific preservation. Non-enveloped vectors such as AAV and adenovirus are generally more stable than enveloped vectors, but they are still vulnerable to aggregation, capsid rupture, genome release, interfacial adsorption, and freeze-thaw damage. AAV products are commonly stored at −80°C for long-term preservation, while optimized liquid formulations may support shorter storage at 2–8°C or 4°C. Buffer composition is critical. An AAV9 formulation containing 5% glycerol and 0.001% Pluronic F68 better preserved viral recovery and infectivity after freeze-thaw cycles than formulations containing only sorbitol or surfactant alone[
150]. Mechanistic studies of AAV8 and AAV9 further showed that freeze-thaw-induced capsid rupture is associated with interfacial stress and excipient devitrification; low concentrations of poloxamer 188, suitable cryoprotectants and reduced buffer concentration can markedly reduce capsid rupture during freezing and thawing[
151]. For lung-directed AAV6.2FF, Pluronic F68 improved thermal and freeze-thaw stability and was associated with higher
in vivo lung epithelial transduction after intranasal administration compared with phosphate-buffered saline (PBS) alone[
152]. The results link buffer, surfactant and cryoprotectant selection to preservation of AAV infectivity before pulmonary administration.
Adenoviral vectors and oncolytic viruses can also be stored as refrigerated liquids or converted into lyophilized products. Liquid chimpanzee adenovirus-vectored vaccines typically require 2–8°C distribution, but lyophilization with sugars, buffers, bulking agents, and surfactants has been developed to reduce refrigeration dependence. Berg et al. evaluated ChAdOx1 and ChAdOx2 liquid and lyophilized formulations, and subsequent formulation optimization achieved ChAdOx lyophilized products with limited infectivity loss during drying and little additional loss over one month at 30°C[
153,
154]. Lyophilization can extend adenoviral-vector stability when the drying cycle, residual moisture, glass-transition temperature, and excipients preserve infectivity.
Enveloped therapeutic viruses, including many lentiviral, retroviral, or vaccinia-based systems, are usually more fragile because envelope proteins and lipid membranes are sensitive to temperature, shear, freeze-thawing, and interfacial adsorption. Lentiviral vectors are commonly stored frozen, often at −80°C, and repeated freeze-thawing is generally avoided because infectivity can decline even when physical particles remain detectable[
155]. Recent work with infectious enveloped viruses showed that spray-drying or lyophilization into glassy polysaccharide matrices can preserve viral titer and improve thermal stability, suggesting that dry-state preservation may become feasible for selected enveloped viral systems when the formulation matrix is carefully designed[
156]. Post-storage infectious titer and transgene expression need to be interpreted with capsid or envelope integrity, aggregation and the genome copy-to-infectivity ratio.
Membrane-bound vesicles: EVs, OMVs, and membrane-derived vesicles
EVs are commonly stored at 4°C for short-term handling and at −80°C for longer-term preservation, but these conditions are not universally protective. Storage in simple PBS can reduce EV recovery, increase aggregation or fusion and alter size, zeta potential, and surface-marker presentation. Gelibter et al. systematically showed that −80°C storage and repeated freeze-thaw cycles reduced EV concentration, increased particle size variability and generated double-positive vesicle populations consistent with fusion-like events[
157]. Görgens et al. further demonstrated that PBS alone can rapidly reduce EV recovery, whereas PBS supplemented with human albumin and trehalose markedly improved short-term and long-term EV preservation at −80°C and protected EVs during freeze-thaw cycles[
158]. Wu et al. compared small EV storage at 4°C, −20°C, and −80°C for up to 28 days and concluded that −80°C was preferable for long-term therapeutic preservation, although storage still affected size distribution, quantity, contents, uptake, and biodistribution[
159]. Together, the data favor minimizing freeze-thaw cycles and using optimized protein- and sugar-containing buffers for long-term storage at −80°C; the appropriate condition still requires product-specific functional validation.
Lyophilization is being explored to reduce EV cold-chain dependence, but freeze-drying without suitable excipients can cause aggregation, membrane deformation, and cargo loss. Trehalose, sucrose, poloxamer 188 and albumin-based stabilizers are among the most commonly used protective excipients. Trenkenschuh et al. showed that EV colloidal stability during freeze-thawing and lyophilization was improved by neutral phosphate buffers, 5% sucrose and low concentrations of poloxamer 188; lyophilized EV formulations maintained particle size and concentration for up to 6 months, while associated enzymatic activity was better preserved during early storage[
160]. Lyu et al. found that corneal stromal stem-cell-derived EVs stored at −80°C or lyophilized with trehalose retained morphology, particle recovery, CD9/CD63/CD81 expression and anti-inflammatory or anti-fibrotic activity better than lyophilized EVs without trehalose[
161]. Milk-derived EVs have also been stabilized by trehalose and tryptophan-containing lyophilized formulations, which preserved vesicle structure and bioactivity during room-temperature storage[
162]. Successful lyophilization requires recovery of vesicle structure and cargo together with target-cell uptake and biological activity after reconstitution.
OMVs can tolerate some storage stresses, but formulation changes may alter the membrane components responsible for immunological potency. Preserving OMVs means preserving vesicle structure and antigenic display while preventing uncontrolled changes in LPS activity or outer membrane protein composition. Experimental studies using EV and OMV model systems showed that freeze-thaw damage was most pronounced in PBS and could be reduced using low-ionic-strength phosphate buffers, sucrose and poloxamer 188; these conditions helped maintain vesicle concentration and size after freeze-thaw stress and lyophilization[
155]. Earlier OMV cargo-stability studies also showed that enzymes packaged inside OMVs retained much higher activity than free enzymes after elevated temperature exposure, repeated freeze-thaw cycles and lyophilization, supporting the idea that the OMV lumen can protect sensitive payloads under storage stress[
163]. For vaccine-like OMVs, cold storage remains important. OMV-based meningococcal vaccine formulations have historically relied on refrigerated or frozen storage, and recent drying approaches such as light-assisted drying in amorphous trehalose matrices have preserved OMV-containing 4CMenB vaccine nanostructure, antigenicity and
in vivo immunogenicity after processing[
164]. Particle integrity must therefore be linked to antigen display, lipid A, or endotoxin activity and immune potency after storage.
Storage data for cell membrane-derived vesicles remain less systematic than those for EVs and OMVs. Their function depends on membrane orientation, receptor and glycan retention and preservation of source-cell recognition signals, yet many protocols are extrapolated from EVs or membrane-coated synthetic particles. Such extrapolation is uncertain because vesiculation and purification can alter membrane sidedness, protein density, and aggregation. Formulation studies need direct measurements of vesicle recovery, membrane orientation, target binding, and biological activity after storage.
Aerosolization and pulmonary delivery
Aerosolization introduces stresses that are not captured by conventional storage studies. Shear, air-liquid interfaces, impaction, heating, osmotic shifts, and device adsorption can reduce biological activity even when emitted dose and aerodynamic size remain acceptable. Aerosol performance must therefore be assessed together with platform-specific potency. The following sections compare these requirements in living, infectious, and vesicular systems.
Quantitative aerosol characterization remains incomplete for many biological platforms. Although some studies reported nebulizer type, aerodynamic particle size, fine-particle fraction, or functional recovery after aerosolization, many provided only the nominal administered dose and therapeutic outcome. Emitted dose, regional deposition, and post-aerosol potency were often not reported (Table 4). Therefore, missing parameters are identified as not reported rather than estimated, and local instillation studies are distinguished from true aerosol inhalation.
Living systems: bacteria, mammalian cells, and microalgae
Evidence remains uneven. Bacteria and microalgae have emerging inhaled and dry-powder examples, while intact mammalian cells are still delivered mainly by intravenous infusion, intratracheal instillation, or local spraying rather than conventional nebulization.
For bacterial systems, the central challenge is maintaining viable CFU while preventing rapid macrophage clearance, environmental release and uncontrolled lung colonization. Most bacteria-based pulmonary therapy studies still rely on intratracheal administration rather than validated aerosol delivery, indicating that aerosol engineering remains underdeveloped for live bacterial therapeutics. An inhalable example is the
Bdellovibrio bacteriovorus-loaded poly lactic-co-glycolic acid (PLGA) large porous microsphere developed for drug-resistant
Pseudomonas aeruginosa pneumonia. In this design, the large porous microsphere was engineered as a dry-powder carrier and a protective “safe house” for predatory bacteria. Its large geometric size reduced alveolar macrophage phagocytosis, whereas the porous structure enabled pulmonary inhalation and local release of viable
B. bacteriovorus at the infected lung site[
165]. The study reported a respirable dry powder, protection of viable bacteria, and antibacterial activity after lung deposition. Other bacterial microrobots and drug-loaded constructs were administered mainly by intratracheal instillation[
94,
95] and therefore do not establish compatibility with non-invasive inhalation devices.
Evidence for aerosolized mammalian-cell carriers is limited. Conventional nebulization of intact cells is difficult because mammalian cells are large, fragile, deformable, and prone to aggregation, and because airway obstruction or loss of phenotype would be unacceptable for clinical translation. Most pulmonary cell therapies therefore use intravenous delivery, which exploits first-pass pulmonary trapping, or direct local routes such as intratracheal instillation and bronchoscopy-guided administration. Available aerosol-related studies are mainly feasibility studies rather than mature therapeutic aerosol systems. Thiebes et al. evaluated endoscopic atomization of MSCs using air and pressure atomization devices and found that immediate survival remained above 90%, with preserved MSC morphology, surface marker expression and trilineage differentiation capacity after spraying[
76]. The result supports bronchoscopic spraying but cannot be extrapolated to patient-controlled inhalation through a conventional nebulizer. Castillo Aleman et al. compared compressor, ultrasonic and mesh nebulization of human peripheral blood-derived stem-cell populations
in vitro. Compressor nebulization better preserved selected cell counts and viability, whereas mesh nebulization caused greater loss of viability in their setting, highlighting that device choice strongly affects cell survival[
75].
Some microalgal species are more amenable to aerosolization than mammalian cells because they can be encapsulated within inhalable droplets and retain self-propulsion after nebulization. However, the relevant biological potency is not simply viability; the algae must continue to swim and retain surface-conjugated cargo after aerosol generation. Early microrobot studies in
P. aeruginosa pneumonia and lung metastasis used intratracheal delivery and therefore did not establish non-invasive aerosol administration[
24,
166]. Li et al. subsequently used a nebulizer to encapsulate picoeukaryotic algae microrobots within small particles for inhalation. The algae retained motility after nebulization, achieved homogeneous lung distribution and remained in mouse lungs for more than five days. When functionalized with platelet membrane-coated vancomycin-loaded nanoparticles, the inhaled algae microrobots improved treatment of acute MRSA pneumonia[
25]. The study links droplet encapsulation and inhaled dose to post-nebulization motility and therapeutic function. A hybrid consortium using motile
Synechococcus to carry PEGylated oncolytic adenoviruses extends the approach to a biological payload[
111]. Both algal motility and viral infectivity must be measured after inhalation to establish performance of the combined system.
Infectious systems: bacteriophages and therapeutic viruses
Bacteriophages have reached early clinical testing as pulmonary aerosols. Their main aerosol-specific problems are loss of PFU during nebulization or drying, altered phage-cocktail ratios, humidity-sensitive powder instability, and reduced bacterial adsorption after aerosolization. Liquid nebulization has already reached early clinical testing. Weiner et al. developed BX004-A, a nebulized three-phage cocktail targeting
Pseudomonas aeruginosa in cystic fibrosis, and a first-in-human phase 1b/2a trial showed acceptable safety, lower-respiratory-tract delivery, and preliminary reduction of sputum bacterial burden[
65]. Chan et al. used a personalized nebulized phage strategy in adults with cystic fibrosis and multidrug-resistant or pan-drug-resistant
P. aeruginosa; the selected single phages or cocktails were delivered by nebulization without adverse events and reduced sputum bacterial density while preserving the sputum microbiome[
64]. The clinical data make host range, stability during nebulization, and lower-airway PFU central to dose interpretation alongside lung deposition.
Dry-powder phage engineering provides an alternative to liquid nebulization and may improve storage, portability, and patient handling. The excipients must protect phage structure during drying while producing respirable particles after dispersion. Lin et al. co-spray dried phage PEV20 with ciprofloxacin using lactose and L-leucine, generating inhalable combination powders with high fine particle fractions and preserved antibacterial synergy after dispersion through dry-powder inhalers[
167]. The same phage-antibiotic powder concept was tested in a mouse lung infection model, where insufflated PEV20-ciprofloxacin powder reduced pulmonary
P. aeruginosa burden by nearly six log
10 CFU and reduced infection-associated inflammation[
168]. The formulation linked PFU preservation and powder dispersion with co-deposition of phage and antibiotic and retained antibacterial synergy after pulmonary delivery. Dose-response studies of inhaled PEV31 further showed that delivered PFU influences lung phage replication, resistant subpopulation emergence, and inflammatory cytokine suppression, emphasizing that aerosolized phage therapy requires quantitative control of biologically active dose rather than nominal formulation dose alone[
169].
Therapeutic viruses face a different set of aerosol constraints. Non-enveloped vectors such as AAV and adenovirus are generally more compatible with aerosol delivery than enveloped viruses, but they can still lose infectious titer through capsid damage, aggregation, adsorption to device surfaces, or neutralization in airway fluids. AAV-based pulmonary gene therapy has therefore focused on capsid engineering and compact transgene design that support functional delivery after aerosol administration. Calton et al. developed 4D-710, an aerosolized cystic fibrosis gene therapy using the directed-evolution-derived A101 AAV capsid and a shortened CFTR-ΔR transgene. The A101 capsid was selected for efficient airway epithelial transduction, including in the presence of pre-existing neutralizing antibodies, after aerosolized administration to nonhuman primates, while the compact CFTR cassette enabled packaging and functional rescue in cystic fibrosis airway epithelial cultures[
66]. The design combines capsid and transgene engineering; functional epithelial transduction after inhalation remains the decisive aerosol endpoint.
Adenoviral vectors provide another relatively advanced aerosol-compatible viral platform, especially for respiratory vaccination. Aerosolized Ad5-nCoV was evaluated in a phase 1 clinical trial and showed acceptable safety and immunogenicity after inhaled delivery[
170]. More recently, inhaled multi-antigen adenoviral COVID-19 vaccines were delivered as aerosols to humans and induced lung mucosal T-cell, trained innate and antibody immunity, with chimpanzee adenoviral vectors outperforming human Ad5 vectors in that setting[
171]. These vaccine studies do not establish treatment of established lung disease, but they provide human data on aerosol dose, vector stability, mucosal deposition, and local immune activation. For oncolytic viruses, true aerosol-engineered pulmonary delivery remains much less mature. A hybrid consortium used motile
Synechococcus to carry PEGylated oncolytic adenoviruses into lung tumors and improved local distribution[
111]. The result is specific to a combined experimental system and does not establish a general aerosol platform for oncolytic viruses. Enveloped viral vectors, including many lentiviral or retroviral systems, remain particularly challenging for aerosolization because their lipid envelopes and entry glycoproteins are sensitive to shear, drying, and interfacial stress.
Vesicular systems: EVs, OMVs, and membrane-derived vesicles
Vesicles’ aerodynamic behavior is determined by the micron-scale droplets or powder particles that carry them. Post-aerosol evaluation should therefore link conventional aerosol measurements to vesicle recovery and functional cargo delivery.
EVs currently provide the strongest evidence among vesicular biological aerosols. Their main aerosol-specific challenges are maintaining membrane integrity and RNA/protein cargo activity after nebulization, while achieving lung deposition and uptake by relevant pulmonary cells. Popowski et al. showed that lung-derived EVs outperformed liposomes and HEK293-derived EVs as inhaled carriers for mRNA and protein delivery. After jet nebulization, lung-derived EVs showed better distribution and retention in bronchioles and lung parenchyma, supporting donor-cell selection as an aerosol-relevant engineering variable[
121]. Han et al. further developed a vibrating mesh nebulization method for small EVs loaded with small RNAs. Nebulized sEVs were delivered mainly to the lung and taken up by macrophages and airway epithelial cells, and siMyd88-loaded sEVs reduced LPS-induced lung injury in mice[
47]. Device choice and vesicle source influenced post-nebulization distribution and functional RNA delivery, making both variables part of aerosol design.
EV aerosols have also been extended from delivery feasibility to disease treatment. Liu et al. reported inhalable EVs carrying IL-12 mRNA for lung cancer immunotherapy. Inhaled IL-12 mRNA-loaded EVs enabled local cytokine expression in lung tumors, promoted antitumor immunity and reduced systemic toxicity associated with cytokine therapy[
122]. More recently, inhalable hybrid cellular vesicles combining neutrophil membrane-derived vesicles, PD-L1-overexpressing MSC EVs and resveratrol were administered by nebulization to treat lung ischemia-reperfusion injury and MRSA pneumonia. After nebulization, the hybrid composition retained inflammatory targeting and immunomodulatory activity[
172]. Emitted dose must therefore be related to vesicle recovery, cargo retention, aggregation, cell-specific uptake, and downstream activity.
OMVs are also attractive for pulmonary aerosol delivery because they are stable bacterial membrane vesicles with intrinsic adjuvant activity. Their aerosol-specific problem is different from EVs: OMVs must preserve antigen display and immunostimulatory potency while avoiding excessive LPS-driven lung inflammation. Miao et al. developed an inhalable bacteria-derived membrane vesicle nanovaccine from engineered
E. coli expressing ClyA-OVA257-264. Inhaled BMVax promoted antigen cross-presentation, activated tracheobronchial lymph node immunity and produced stronger protection against lung metastasis than subcutaneous immunization[
126]. Xu et al. used glycine-induced OMVs as pulmonary
in situ vaccines for metastatic lung cancer; preparation-method engineering reduced LPS content, enriched immunogenic outer membrane proteins, and improved safety during pulmonary delivery[
55]. The studies establish pulmonary administration of OMV immunotherapies, but reporting of emitted vesicle dose, aerodynamic size, post-nebulization integrity, LPS activity, and antigen display remains incomplete.
Cell membrane-derived vesicles and hybrid membrane vesicles are less mature as aerosol platforms. Their therapeutic effects depend on preserving membrane orientation, surface receptors, glycans and immune-recognition signals. Nebulization may disrupt these membrane features or detach loaded drugs. The Res-PD-L1@nmEVs study provides a recent example in which neutrophil membrane-derived vesicles and engineered MSC EVs were integrated into an inhalable hybrid vesicle system and delivered by nebulization, resulting in enhanced pulmonary accumulation and suppression of excessive neutrophil activation in inflammatory lung injury models[
172]. However, compared with EVs, there are fewer systematic studies showing how atomization, droplet formation or dry-powder dispersion affect membrane-derived vesicle orientation and surface functionality. The available evidence is therefore insufficient to determine how aerosolization affects membrane orientation and surface function across membrane-derived vesicle products.
Remaining challenges and translational perspectives
Clinical translation requires the engineered function to remain measurable from product release through pulmonary administration. The main unresolved issues are the definition of potency, control of safety and immune activity, preservation of critical quality attributes during manufacture and formulation, and validation in models that reproduce the intended inhalation route.
Potency metrics and quality control
Nominal mass, particle number, and inhaled dose cannot serve as universal potency measures because the active unit differs among biological platforms. Manufacturing, storage and aerosolization may leave physical counts unchanged while reducing viability, infectivity, target binding, or cargo activity. Release testing therefore needs to pair product identity with a functional assay linked to the mechanism of action.
For living products, relevant measures include viable-cell recovery together with phenotype, secretion, motility or cargo retention. Phage and viral-vector assays should relate particle or genome counts to PFU, infectious titer, host adsorption, or transgene expression. Vesicle products require measures of membrane integrity, cargo retention, and functional uptake, with additional assessment of lipid A or endotoxin activity for OMVs. These principles are consistent with mechanism-based potency expectations for cell and gene therapies and with MISEV2023 guidance for EV characterization[
10,
173]. AAV quality attributes require interpretation alongside infectivity or transgene expression[
174,
175], while phage activity also depends on adsorption, bacterial susceptibility, resistance, and host clearance[
176]. The resulting specification should describe the biologically active dose after aerosolization, not only the amount loaded into the delivery device.
Safety and immune controllability
Delivering therapeutics directly to the lungs can effectively reduce systemic drug exposure, but bioactive substances will directly come into contact with the highly sensitive mucosal barrier at the same time. Therefore, to ensure patient safety, it is necessary to regulate the therapeutic kinetics of the drugs. Specifically, it is about balancing the dosage and exposure time of the drug administration with the natural ability of the lungs to clear these foreign substances.
Each biological platform presents a completely distinct toxicity profile, so every platform demands tailored mitigation strategies. For live bacterial platforms, investigators must prioritize strict genetic containment to prevent residual virulence and unwanted horizontal gene transfer[
77]. Conversely, the pulmonary administration of therapeutic mammalian cells introduces significant physical and biological hurdles, notably the risk of tissue trapping, aggregation, and unpredictable shifts in their immunomodulatory behavior[
177,
178]. When utilizing bacteriophages, safety concerns shift heavily toward manufacturing impurities, alongside the intense inflammatory cascades triggered by rapid bacterial lysis and subsequent microbiome disruption[
176,
179]. Mammalian viral vectors face an entirely different set of immunological obstacles, particularly the rapid activation of complement pathways and the generation of neutralizing antibodies that severely restrict repeated dosing protocols[
180,
181]. Finally, while EVs risk delivering their cargo to unintended cell populations, bacterial OMVs frequently provoke severe local inflammation driven directly by their innate lipopolysaccharide content[
10,
130].
Manufacturing, formulation, and scalability
Traditional drugs have a single chemical structure and thus a single function. In contrast, the advantage of biological agents is their complex functional characteristics. Regardless of the delivery carrier used, researchers always face multiple challenges, such as ensuring batch consistency, precisely removing impurities generated during processing, and maintaining the drug efficacy during the process from initial cultivation to final inhalation administration.
The differences in production modes will lead to different production bottlenecks. For live bacterial preparations, the production process must strictly control the source of the strain, the continuous transmission history, genetic stability, and reduce environmental risks[
182]. Culturing therapeutic cells from mammals faces significant donor variability. The problem of production volume limitation of cell populations has been solved by volume expansion in bioreactors, but researchers still need to ensure that these expanded cell populations maintain their original biological activity[
183,
184]. The production process of bacteriophages requires sequential large-scale virus amplification and strict extraction of host cell debris, residual nucleic acids, and endotoxins. This purification process becomes complex due to the strict time requirements of personalized medicine[
185–
187]. AAV vectors have problems of shell volume accumulation and unwanted particle aggregation in clinical production[
188,
189]. Separating EVs or bacterial vesicles requires highly scalable extraction methods to ensure the concentration of these particles without damaging their fragile lipid boundaries or losing the substances encapsulated[
190,
191]. Ultimately, establishing robust critical quality attributes during manufacturing is only meaningful if those parameters reliably predict the platform’s functional survival following aerosolization and subsequent deposition within the lung.
Pulmonary delivery models and clinical translation
The initial assessment of respiratory biological therapy is mainly based on two methods: direct intravenous infusion of fluids or nasal drops to establish the basic evidence of the therapeutic concept. However, these administration methods cannot simulate the complex aerodynamic characteristics of the actual inhalation process in humans. They ignore the complex structure of respiratory tract, the natural breathing rhythm, and the continuous mucosal clearance movement[
192,
193]. This distinction is important because clinical aerosolization can generate substantial shear that damages carrier integrity, infectivity, or immunomodulatory function before deep-lung deposition. Bridging this translational gap requires a highly robust analytical pipeline. Researchers must explicitly link the specific microscopic droplet fractions generated by medical devices to actual regional tissue deposition and sustained biological potency following aerosol exposure[
193,
194]. Because no solitary experimental setup can faithfully mimic the entire human respiratory system, investigators must actively integrate multiple distinct methodologies.
In inhalation therapy or pulmonary delivery studies, the reliability of the treatment strategy needs to be verified step by step through multiple levels of models. Firstly, computer simulations and traditional experiments can be used to observe the movement and deposition location of aerosol particles in the lungs. Then, microfluidic “lung chips” and lung epithelial cell models cultured at the gas-liquid interface are used to determine whether the drug can pass through the mucosal barrier and whether it will trigger local immune responses. At the same time, using fully isolated lung tissue can more realistically supplement the impact of lung structure on drug deposition. Finally, it is necessary to use mammalian models with clear disease characteristics to further confirm the distribution of the treatment material in the body and its clinical efficacy. The engineering of active biological materials and their derivatives for the diagnosis and treatment of pulmonary diseases must go through steps such as simulation, cell experiments, lung tissue experiments, and animal experiments. Only when all processes are qualified can this approach ensure its possibility as a clinical medication[
195,
196].
Conclusion
Biogenic therapeutics offer features that are unobtainable with conventional synthetic medications. Unlike passive and inactive molecules, they can actively maneuver heavy mucus environments, facilitate cellular interactions, and carry out targeted treatment. However, the clinical potential of this technology is restricted by the severe survival requirements due to the physical stresses involved in large-scale production, as well as the aerodynamic stress associated with respiratory delivery.
Despite this promise, the level of evidence differs greatly among these platforms. Nebulized bacteriophage therapy and aerosolized adenoviral vaccination have entered early human studies. In contrast, microalgal microrobots, engineered cell carriers, OMV immunotherapies, EVs, and hybrid bio-derived vesicles remain mainly at the small-animal stage. Intranasal or intratracheal administration can show local pulmonary activity, but it does not confirm compatibility with patient-relevant aerosol inhalation. Future studies should use clinically relevant inhalation devices and report aerosol properties, delivered dose, biological activity after aerosolization, and long-term safety.
Therefore, advancing these potential therapies toward clinical application will require additional steps beyond conventional testing protocols, including specialized assays for each platform, strict immunomodulatory measures, and preclinical models that accurately replicate the physics of human inhalation. Success will depend not only on intrinsic biological activity, but also on retaining functional potency through formulation and aerosol delivery, achieving effective pulmonary deposition, and withstanding clearance mechanisms in diseased lungs.
The Author(s) 2026. This article is published by Higher Education Press at journal.hep.com.cn.
PDF
(4748KB)
