Cell-derived vesicles loaded with siRNA cargo for targeted gene silencing in cancer
Every cell in the body is continuously secreting tiny lipid-membrane vesicles called exosomes. Unlike synthetic nanoparticles which must be engineered to avoid rapid immune clearance, exosomes evolved over billions of years to navigate biological environments safely — making them compelling as natural drug delivery vehicles.
Exosomes are formed through a specialized sub-compartment of the endosomal system:
1. Early endosome formation: • Receptor-ligand complexes (e.g., EGFR after EGF stimulation) are internalized from the plasma membrane via clathrin-coated vesicles (70–100 nm) • Vesicles acidify (pH 6.0–6.5) and fuse to form early endosomes (300–500 nm) • Rab5 GTPase governs this step
2. Late endosome / Multivesicular Body (MVB) formation: • The ESCRT (Endosomal Sorting Complexes Required for Transport) machinery recognizes ubiquitinated cargo on the early endosome membrane • ESCRT-0 (Hrs/STAM) captures monoubiquitinated proteins • ESCRT-I (TSG101/VPS28/VPS37) recruits ESCRT-II • ESCRT-II nucleates inward membrane budding → intraluminal vesicles (ILVs) form • ESCRT-III (CHMP proteins) drives membrane scission • Alternatively: ceramide/sphingomyelinase pathway forms ILVs independently of ESCRT • MVB interior pH drops to 4.5–5.0 (late endosome) • Rab7 governs MVB maturation
3. MVB-plasma membrane fusion and exosome secretion: • Rab27a/Rab35 regulate MVB trafficking to the plasma membrane • SNARE complex (VAMP7/VAMP3 + syntaxin 4) drives membrane fusion • ILVs are released to extracellular space as exosomes • Secretion is enhanced by: hypoxia (tumor microenvironment), low intracellular Ca²⁺, ceramide, neutral sphingomyelinase activation
The ESCRT machinery provides exquisite control over what cargo is sorted into exosomes. Cancer cells have hijacked this system — tumor exosomes carry oncoproteins, immune checkpoint ligands, and drug resistance factors, enabling cross-talk with stromal cells and immune reprogramming. Therapeutic exosomes can "reverse-hijack" this system to deliver anti-cancer payloads along the same natural routes.
Exosomes are not empty lipid shells — they carry a rich assortment of biomolecules:
Lipid bilayer membrane composition: • Enriched in cholesterol (unlike the plasma membrane, MVB-derived membranes are cholesterol-rich) • High sphingomyelin content → liquid-ordered "raft" microdomains • Phosphatidylserine on outer leaflet (unlike cells, where PS is inner-leaflet only) • Lysophospholipids (generate "cone" geometry favoring curvature)
Transmembrane protein cargo: • Tetraspanins: CD63, CD9, CD81 — canonical exosome markers; function in cell adhesion, fusion, and receptor scaffolding • MHC-I and MHC-II: immune antigen presentation even outside a cell • HSPs (Hsp70, Hsp90): chaperones enriched in tumor exosomes • Integrins: determine organ tropism (α6β4 → lung; αvβ5 → liver)
Luminal RNA cargo: • miRNA: ~800–3000 copies per vesicle; typically 22 nucleotides; post-transcriptional gene regulation • mRNA: functional mRNA has been translated in recipient cells (Valadi et al. Nature 2007) • lncRNA, circRNA: regulatory non-coding species increasingly recognized as intercellular signals • No ribosomal RNA or nuclear RNA → not a passive sampling of cell RNA; active sorting
Luminal protein cargo: • Metabolic enzymes (GAPDH, enolase), signaling proteins (Ras, ERK), ESCRT components
The therapeutic properties of exosomes depend critically on their cellular origin:
Mesenchymal Stem Cells (MSCs) — most commonly used for therapeutics: • Natural homing to sites of inflammation and injury → loaded exosomes naturally accumulate at disease sites without additional targeting • Large-scale production capability: MSCs can be expanded to billions in bioreactors • Low immunogenicity: MSC-derived exosomes can cross allogeneic barriers — the same batch can treat multiple patients • GMP manufacturing established: several companies (Astellia, Evox Therapeutics, Codiak BioSciences) have GMP MSC exosome processes • Naturally carry immunosuppressive cargo (TGF-β, IL-10, IDO) — beneficial for autoimmune applications but may be a concern for cancer therapy where immune activation is desired
Dendritic Cells: • Loaded with tumor antigens → exosomes carry MHC-I:peptide complexes → activate CD8+ T cells even after cell death • Anosys (France) conducted Phase I/II trials: DC-derived exosomes loaded with MAGE tumor peptides → modest T-cell activation
Platelet-derived exosomes: • Abundant from blood processing; carry coagulation factors; used in wound healing and bone regeneration
Engineered exosomes from HEK293T: • Overexpress a specific protein-cargo fusion in the producer cells → exosomes "pre-loaded" with the protein • Codiak BioSciences: CD63-PTGFRN scaffold for deterministic cargo loading into exosome membrane
The challenge of RNA delivery has been one of the central problems of modern medicine. siRNA molecules are rapidly degraded by serum nucleases, cannot cross cell membranes unaided, and trigger immune responses when delivered systemically. Exosomal encapsulation provides natural protection, leveraging the same membrane that the cell uses to package and protect its own RNA.
Getting siRNA into exosomes without destroying the vesicle structure or cargo is technically challenging. Four methods are in use:
1. Electroporation (gold standard for clinical development): • Exosome suspension + siRNA mixed in EP buffer • Pulse: 400 V/cm, 100–200 μs pulse duration; or 2.0 kV 5ms (exponential decay) • Physical principle: brief electrical field creates transient pores in lipid bilayer • siRNA (<7 nm) enters through pores before they reseal • Efficiency: 15–40%; loss of siRNA activity potential (aggregation, siRNA-to-external membrane sticking) • Best validated for clinical translation; scalable
2. Chemical co-incubation (saponin-assisted): • Saponin (membrane-permeabilizing detergent at sub-lytic concentrations) + siRNA co-incubated with exosomes at 37°C for 1–2h • No heat or electricity; maintains vesicle integrity better than EP in some studies • Risk: saponin cannot be fully removed → cytotoxicity at high concentrations
3. Hydrophobic modification: • Cholesterol-conjugated siRNA (chol-siRNA) or tocopherol-siRNA self-inserts into lipid bilayer • Loading is co-incubation overnight; incorporated into outer membrane leaflet • High loading efficiency (>80%); concerns about siRNA release from membrane (may require endolysosomal processing)
4. Sonication / extrusion: • High-frequency ultrasound (40 kHz, 30s) temporarily disrupts membrane → siRNA mixes into interior • Or: mix cell membrane + siRNA and force through polycarbonate membranes (100 nm pores) • Simplest; risk of vesicle destruction and aggregation
The ideal loading method preserves exosome structure, achieves high siRNA encapsulation, ensures endosomal release, and is scalable to GMP manufacturing. No single method currently satisfies all four. Electroporation leads in clinical translation, but modified lipid-insertion methods may surpass it in efficiency for specific cargo types.
Once delivered to the cytoplasm, siRNA enters the RNA interference (RNAi) pathway — one of the most ancient and conserved gene regulation mechanisms in eukaryotes:
Molecular mechanism: 1. siRNA enters cytoplasm (21–23 nt double-stranded RNA; 2-nt 3′ overhangs conform to Dicer cleavage product recognition) 2. Dicer (an RNase III enzyme) may further process the duplex; Hsp70/Hsp90 facilitate loading into RISC 3. Argonaute 2 (AGO2) is the catalytic core of RISC (RNA-Induced Silencing Complex) 4. Strand selection: the antisense (guide) strand with lower 5′ thermodynamic stability is preferentially retained; sense (passenger) strand is cleaved by AGO2 and discarded 5. Guide strand:mRNA hybridization: AGO2 scans mRNAs for sequences complementary to the guide strand 6. Perfect complementarity (siRNA): AGO2 cleaves the mRNA between positions 10 and 11 in the guide:mRNA duplex (slicer activity) 7. RISC is catalytic — one siRNA molecule can silence hundreds of mRNA molecules 8. Durability: siRNA activity lasts 5–14 days (diluted by cell division); for non-dividing cells much longer
Design considerations: • GC content: 30–50% optimal • Avoid thermodynamically stable internal hairpins • Avoid off-target seed region matching: minimize 7-mer matches of positions 2–8 to non-target mRNAs • Chemical modifications: 2′-OMe, 2′-F at specific positions → nuclease resistance + reduced immunostimulation (TLR7/8 recognition)
The choice of siRNA target is as critical as the delivery method. Validated cancer targets:
1. KRAS (Kirsten RAS): • Mutated in 30% of all cancers; >90% of pancreatic cancers • "Undruggable" by conventional inhibitors for decades (no binding pocket) • First approved KRAS inhibitor (sotorasib, KRASG12C) only covers ~13% of KRAS mutations • siRNA can silence ALL KRAS mutants regardless of amino acid change • Arrowhead: ARC-EX001 — exosome-delivered KRAS siRNA in pancreatic cancer; preclinical 85% tumor regression
2. PLK1 (Polo-like kinase 1): • Mitotic kinase essential for cell division • Highly expressed in many tumor types, low in most normal tissues • Alnylam: ALN-PLK — lipid nanoparticle siRNA; showed tumor regression in Phase I • Exosome delivery could widen therapeutic window by avoiding LNP liver tropism
3. BCL-2/BCL-XL: • Anti-apoptotic proteins; overexpressed in many cancers; confer drug resistance • siRNA knockdown sensitizes resistant tumors to chemotherapy • Combination: exosome-siRNA-BCL2 + paclitaxel → 10× enhanced killing vs. either alone
4. HIF-1α: • Master hypoxia regulator; drives VEGF (angiogenesis), glycolysis genes (Warburg effect), drug resistance • Tumors become resistant to anti-VEGF therapy partly via HIF-1α upregulation • siRNA silencing of HIF-1α: reduces tumor vascularity + re-sensitizes to chemotherapy
Wild-type exosomes accumulate predominantly in the liver and spleen after systemic injection — cleared by resident macrophages before reaching tumors. Surface engineering transforms these natural vesicles into precision-guided nanocraft that can find specific cancer cells while evading the immune system long enough to deliver their payload.
Three classes of targeting moieties are conjugated to exosome surfaces:
1. Peptides (preferred for clinical translation): • Small (8–20 amino acids), chemically synthesized, low immunogenicity • GE11 peptide (YHWYGYTPQNVI, 1340 Da): binds EGFR ectodomain with Kd = 22 nM; non-agonist (unlike EGF); derived from phage display against EGFR extracellular domain III • RGD (Arg-Gly-Asp): targets αvβ3 integrins on angiogenic endothelium and many cancer cells; well-validated for tumor targeting • iRGD (CRGDKGPDC): bicyclic peptide — first binds integrins, is cleaved by tumor protease to expose CRGDK, which binds NRP-1 receptor → enhanced deep tissue penetration • cRGDyK: cyclic locked RGD with 10× higher integrin affinity than linear RGD • Coupling chemistry: peptide-NH₂ reacts with NHS-activated DSPE-PEG; conjugate is mixed with exosomes → inserts spontaneously into bilayer via lipid anchor
2. Monoclonal antibodies / Fab fragments: • Highest selectivity (sub-nanomolar to picomolar Kd) • Anti-HER2 (Trastuzumab) Fab coupled via azide-DBCO click chemistry • Large size (150 kDa IgG vs. 1–2 kDa peptide) may sterically block membrane fusion; prefer Fab (50 kDa) or nanobody (15 kDa)
3. Aptamers (nucleic acid ligands): • Anti-EpCAM, anti-PSMA, anti-AS1411 G-quadruplex (nucleolin target in cancer cells) • SELEX-derived single-stranded DNA/RNA sequences; Kd in nM range • Conjugated via 5′-cholesterol modification → membrane insertion; or streptavidin-biotin to DSPE-PEG-biotin
The iRGD bicyclic peptide demonstrated a remarkable "co-delivery" effect: co-injection of free iRGD with any nanoparticle (liposomes, [], even small molecules) enhances their tumor accumulation 3–4× by activating NRP-1-mediated transcytosis across the tumor endothelium. This tumor-penetrating capability overcomes the fundamental barrier separating blood vessels from tumor parenchyma.
The dominant fate of systemically injected nanoparticles is hepatic/splenic clearance by monocytes and Kupffer cells (resident liver macrophages) — the mononuclear phagocyte system (MPS). Exosome strategies to evade this:
1. CD47 "don't eat me" signal: • CD47 is a transmembrane protein expressed on all nucleated cells and platelets • Binds SIRPα (signal regulatory protein α) on macrophages → activates immunoreceptor tyrosine-based inhibitory motifs (ITIMs) → prevents phagocytosis • Exosomes from CD47-overexpressing donor cells carry more CD47 → extend circulation • Alternatively: CD47-ectodomain peptide mimetic ("CD47 peptide") conjugated to exosome surface • Effect: 2–3× longer plasma half-life; 4× less liver uptake in mice
2. PEGylation: • Short-chain polyethylene glycol (MW 1–5 kDa) grafted to DSPE-PEG lipids that insert into exosome membrane • Hydrophilic PEG layer sterically blocks opsonins (IgG, complement C3b) from adsorbing onto surface • Trade-off: dense PEG may reduce receptor-targeting efficiency (shields targeting ligands) • Optimal: 2–5% PEG density; targeting ligands conjugated to distal PEG ends
3. Surface protein removal: • Phosphatidylserine (PS) on outer leaflet triggers complement clearance via C1q binding and Protein S/Gas6 recognition by TAM receptors on macrophages • Exosome PS can be masked by annexin V decoration, Mer kinase inhibition, or PS-specific nanobody coating • Reduces complement activation and macrophage uptake
Rather than post-isolation chemical conjugation, exosome surface molecules can be encoded genetically in the producer cell:
"Exosome display" strategy: • Fuse your targeting protein/peptide to the extracellular domain of a tetraspanin (CD63, CD9, CD81) or LAMP2b (lysosome-associated membrane protein 2b) • Transfect producer cells with the fusion construct • All secreted exosomes carry the fusion protein on their surface — no downstream chemistry required • Validated: LAMP2b-RVG (rabies virus glycoprotein peptide) exosomes delivered siRNA across the blood-brain barrier (Alvarez-Erviti et al. Nature Biotechnology 2011 — landmark paper in the field)
Advantages: • Consistent, stoichiometric display — every vesicle has the targeting molecule • No chemical modification steps → simpler GMP manufacturing • Protein is transmembrane: more stable than lipid-inserted peptides • Can combine multiple surface modifications (e.g., targeting + imaging + stealth) by co-expressing multiple fusion constructs
Limitations: • New construct validation required for each target — more upfront biology • Expression levels vary by cell line and passage number • Regulatory agencies may require characterization of the entire fusion protein construct as a novel biologic
Codiak BioSciences (acquired 2023) developed "engEx" platform: PTGFRN (scaffold protein abundant in exosomes) used as display scaffold for antibody fragments, cytokines, and small molecules
After intravenous injection, an engineered exosome must survive a gauntlet of biological checkpoints — serum nucleases, complement activation, macrophage surveillance in liver and spleen — before reaching the tumor. Understanding exosome pharmacokinetics allows rational design of delivery strategies that maximize tumor accumulation.
Exosome biodistribution follows a two-compartment model with rapid distribution phase followed by slower elimination:
Distribution phase (0–30 min): • Rapid uptake by liver Kupffer cells (50–70% of injected dose) • Spleen marginal zone macrophages: 10–20% • Lung interstitial macrophages: 5–10% • Remaining circulating fraction: 10–30% → this population is available for tumor targeting
Elimination phase (30 min–24 h): • Circulating exosomes are cleared by: - Complement pathway: C1q, C3b deposition → opsonization → clearance - Scavenger receptors (SR-A, CD36) on liver sinusoidal endothelial cells - LDL receptor-related protein (LRP-1) — binds apolipoproteins that adsorb onto exosome surface
Factors that extend circulation: • CD47 coating (×2–3 half-life) • PEGylation (×2–5 half-life) • Exosome size < 100 nm (smaller vesicles evade spleen mechanical filtration) • Surface charge optimization: -20 to -30 mV reduces non-specific protein adsorption vs. strongly negative (-50 mV) or positive (+charge) particles
Tracking methodology: • Near-infrared fluorescent lipid dye (DiR, DiI) for in vivo optical imaging • Radiolabeled: ¹²⁴I-NHS or ⁶⁴Cu-DOTA conjugated to exosome surface → PET imaging for quantitative biodistribution (clinical-grade) • Bioluminescence: luciferase mRNA cargo → recipient cell luciferase expression reports delivery
The Enhanced Permeability and Retention (EPR) effect has been the dominant rationale for nanoparticle cancer drug delivery for decades:
Mechanism: • Tumor vasculature is structurally abnormal: endothelial cells grow rapidly and incompletely → inter-endothelial gaps of 100–780 nm, vs. 5–10 nm in normal vasculature • Large fenestrations allow nanoparticles (20–200 nm) to extravasate from blood into tumor interstitium • Tumor lymphatics are non-functional → nanoparticles cannot be drained → accumulate → EPR effect
Clinical reality check (sobering): • Retrospective meta-analysis (Wilhelm et al. Nature Reviews Materials 2016): median tumor accumulation across >100 nanoparticle studies = 0.7% of injected dose (ID) • High variability (0.01–4% ID) depending on particle design, tumor model, and tumor perfusion • Human tumors are more heterogeneous and fibrotic than murine models → EPR effect is weaker in patients • EPR works best for: highly vascular tumors (hemangiomas, HCC) and small primary tumors (<100 mm³); poorly for desmoplastic tumors (pancreatic) and necrotic cores
Active targeting synergy with EPR: • Active targeting (surface peptide / antibody) does NOT increase total nanoparticle accumulation in tumors substantially • Active targeting increases intracellular uptake by tumor cells after EPR-driven extravasation → better endosomal escape → improved silencing per particle • Combined EPR + active targeting → 10–30× better knockdown efficiency vs. untargeted particles
IV injection is not the only delivery route for exosome therapeutics:
Inhaled / Intranasal: • For lung cancer, COPD, COVID-19 pulmonary disease: nebulized exosomes (1–5 μm aerosol droplets containing exosome suspension) deposit directly on airway epithelium • Avoids first-pass liver clearance entirely • Exosome-mediated miRNA delivery to alveolar epithelium demonstrated in IPF (idiopathic pulmonary fibrosis) mouse models • Safety advantage: local delivery reduces systemic exposure
Intratumoral injection: • Particularly powerful for solid tumors accessible by ultrasound/CT-guided needle • Eliminates circulation and EPR as barriers • Loading the TME directly with exosomes carrying immune activators (STING agonists, IL-12 mRNA) • Being tested alongside CAR-T therapy to transform "cold" tumors into "hot" immunogenic tumors
Oral delivery for gut tumors: • Exosomes from bovine milk are acid-stable and survive GI tract processing • Bovine milk exosomes loaded with curcumin or siRNA: oral bioavailability in rodents of 20–50% • Clinical interest for colorectal cancer and inflammatory bowel disease • Regulatory pathway: FDA is evaluating milk-derived exosomes under GRAS framework
Local infusion (intracerebral): • Convection-enhanced delivery bypasses blood-brain barrier entirely • Exosome-RVG (brain-targeting peptide) loaded with siRNA for GBM or Huntington's disease
Delivery of siRNA to the tumor cell surface is only half the battle. The cargo must escape the endosome before it is degraded in the lysosome. The "endosomal escape" step is the rate-limiting barrier for all nucleic acid therapeutics — the reason why hundreds of nanoparticle formulations fail in vivo despite impressive in vitro results.
The endosomal escape problem arises from cell biology's own safety system:
1. Endocytosis pathways for exosome uptake: • Clathrin-mediated endocytosis: most common for 100–200 nm particles; receptor-ligand specific • Macropinocytosis: non-specific, large vesicle (0.2–5 μm); predominates in highly active tumor cells • Caveolae: lipid raft-dependent, caveolin-1 involved; may bypass lysosomal degradation • Direct membrane fusion: exosome-cell membrane fusion (similar to virus entry) → direct cytoplasmic delivery; most efficient but least common
2. Endolysosomal acidification: • Early endosome: pH 6.0–6.5 • Late endosome / MVB: pH 5.0–5.5 • Lysosome: pH 4.5–5.0 • Acidic pH + lysosomal nucleases (RNase H, RNase A) degrade unprotected siRNA within 30–60 minutes
3. Why only 1–5% escape: • Endosome membrane is actively maintained by V-ATPase proton pump and intact LAMP1/2 proteins • Membrane fusion requires energy + specific SNARE proteins (not present on synthetic particles) • Exosomes carry DOPE (dioleoylphosphatidylethanolamine): non-bilayer-forming lipid; at low pH, shifts from lamellar to hexagonal phase → disrupts endosomal membrane → allows leakage • But this is partial and inefficient → most cargo degraded
The 1–5% endosomal escape efficiency means that ~95% of delivered siRNA is wasted in lysosomes. This is not catastrophic because each escaped siRNA molecule can silence hundreds of mRNA targets (RISC is catalytic) — but it does mean effective in vivo doses are in the nanomolar range (10–100 pM intracellular). Further improvement of endosomal escape is the single most impactful advancement possible for RNA therapeutics.
Once in the cytoplasm, siRNA is immediately recognized by the RNAi machinery:
RISC assembly — detailed mechanism: 1. siRNA duplex (21 nt guide + 21 nt passenger) is recognized by Hsp70/Hsp90 complex in an ATP-dependent manner 2. The pre-RISC complex (Hsp70/90 + AGO2 + siRNA duplex) forms; Hsp90 opens the AGO2 PAZ and PIWI domains to accept the duplex 3. ATP hydrolysis drives duplex unwinding: the 5′-AGO2-bound guide strand is retained; passenger strand is cleaved by AGO2 slicer activity and discarded 4. Active RISC = AGO2 loaded with single-stranded guide siRNA, in a crescent-shaped clamp conformation
Target recognition and cleavage: 5. AGO2 slides along poly-ribosomes, probing mRNA sequences for complementarity to the guide strand 6. Seed region (positions 2–8 of guide) provides initial binding kinetics (kₒₙ) 7. Perfect complementarity triggers conformational change in the PIWI domain → positions the scissile phosphate between nucleotides 10–11 between the catalytic DEDH tetrad (Asp-Glu-Asp-His) 8. Two-metal-ion mechanism (Mg²⁺) cleaves the 3′-O–P bond → mRNA in two fragments 9. 5′-fragment is degraded by XRN1 exoribonuclease; 3′-fragment degraded by the exosome complex 10. RISC releases cleaved mRNA products and cycles → average RISC complex turns over ~10–200 mRNA molecules before eventual deactivation
Outcome: target oncogene mRNA levels reduced 70–95%; protein levels reduce with a lag equal to protein half-life (hours to days)
Exosome therapeutics have moved from academic discovery to Phase 1 and Phase 2 clinical trials:
Codegenix / Evox Therapeutics: • EVX-101 (MSC-derived exosomes for Duchenne muscular dystrophy): exon-skipping ASO cargo; Phase 1 recruiting • Platform: "ExoSTING" — exosomes with STING agonist surface display; Phase 1/2 in solid tumors (intratumoral injection)
KI Therapeutics / Arrowhead Pharmaceuticals: • exoASO-STAT6: exosome-delivered antisense oligonucleotide targeting STAT6 (oncogene in M2 tumor macrophages); Phase 1 in solid tumors; exosomes derived from bone marrow MSCs • Preclinical: combination with anti-PD-1 showed complete regression in colorectal cancer model
Amnis Therapeutics: • Meso-derived exosomes loaded with KRAS-G12D siRNA for pancreatic cancer • Rational: tumor macrophages naturally take up exosomes → deliver KRAS siRNA to macrophages within tumor → reprogram M2→M1 phenotype + kill KRAS-dependent tumor cells
Challenges remaining for clinical translation: 1. Scalable, reproducible GMP manufacturing of defined exosome populations 2. Potency assay standardization: what is a "batch-to-batch equivalent" dose? 3. Long-term storage: frozen or lyophilized? 4. Regulatory framework: FDA does not have a dedicated exosome guidance — individual IND required 5. Cost of goods: currently ~$1,000–10,000 per clinical dose vs. target <$100 for commercial viability