Recent developments in messenger RNA-modification techniques have enhanced the therapeutic efficiency of mRNA delivery and its potential for near-term clinical applications, including protein-replacement therapy and vaccination against the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) virus1,2. However, the intrinsic inability and potential immunogenicity of mRNAs require that they be encapsulated within delivery vehicles. Current mRNA-delivery modalities centre on the usage of lipid nanoparticle (LNP) carriers for encapsulation and transport3,4. However, LNPs pose several major challenges, including cytotoxicity, poor biodistribution, lack of target specificity and immunogenicity. These problems may be caused by the requirement for the surface PEGylation (PEG stands for poly(ethylene glycol)) of LNPs to improve their circulatory half-life and to reduce non-specific clearance5,6. Notably, the administration of LNPs in people has been linked to anaphylaxis, hypersensitivity and autoimmune adverse events7,8. Therefore, the identification of mRNA carriers that can overcome some of these LNP-associated challenges would be helpful for the further development of mRNA-based therapeutics.

Extracellular vesicles (EVs), including exosomes and microvesicles, play a major role in the transport of biomolecules and nucleic acids, including mRNAs, within the human body9,10,11. As a result, in recent years, EVs have emerged as promising carriers for nucleic-acid-based therapeutics owing to their intrinsic biocompatibility, their ability to cross physiological barriers and their low immunogenicity12,13. Unlike LNPs, EVs, including exosomes, are endogenously produced by the body’s cells and lead to lower levels of inflammatory responses. Moreover, strategies to cheaply and easily produce large quantities of exosomes have been developed. We previously reported a cellular nanoporation (CNP) method in which transient nanometric pores were created on the surface of source cells to allow for the large-scale loading of full-transcript mRNAs into secreted EVs14. Here, by using a mouse model of acute photoaging that closely mimics the pathophysiological features of aging-damaged skin in humans15, we show the utility of exosome-based COL1A1 mRNA therapy to replace dermal collagen-protein loss as an anti-aging treatment for photoaged skin. To improve the efficiency of mRNA delivery and retention, we also show that the delivery of collagen mRNA via a hyaluronic acid (HA) microneedle (COL1A1-EV MN) patch allows for a more efficient distribution of mRNA in the dermis, resulting in durable collagen-protein engraftment and in an improved treatment of wrinkles in photoaged skin.


Preparation and in vitro delivery of COL1A1 mRNA-containing EVs

Dermal atrophy owing to irreversible loss of collagen is a hallmark of skin aging16,17. Numerous methods have aimed to restore loss of collagen protein in skin, ranging from over the counter and pharmaceutical approaches (antioxidants18,19,20, retinoids21, peptides22,23) to medical devices (that is, laser therapy24 and synthetic dermal fillers25,26). However, none of these existing technologies have been able to achieve long-term endogenous collagen replacement to maintain skin strength, firmness and elasticity over time27,28,29. Stimulating fibroblasts responsible for synthesizing collagen proteins can also be an effective way for short-term control of skin aging30. However, fibroblasts gradually lose their capacity to proliferate and synthesize collagen as they senesce, resulting in challenges for longer-term methods of collagen replacement for anti-aging treatment31. To overcome these limitations, we aimed to replace collagen protein in a photoaging collagen-depletion model via EV-mediated mRNA delivery. To generate EVs loaded with human collagen I alpha I (COL1A1) mRNA, we employed a CNP technique that involved plating a monolayer of neonatal human dermal fibroblasts (nHDFs) on a nanopore surface and nano-transfecting the cells with a COL1A1-GFP plasmid (Fig. 1a and Supplementary Fig. 1a)14. EVs were isolated from culture media the day after transfection. CNP-treated cells were found to have a 10-fold higher EV number per cell as compared with cells treated with standard bulk electroporation (BEP), which was performed using cuvette-style parallel electrodes as previously described32, or untreated nHDFs in culture (Fig. 1b). The EVs produced by each method were characterized by a size distribution peaking at about 150 nm in diameter as determined by nanoparticle tracking analysis (NTA) and 80% of intensity in the mode peak by dynamic light scattering (DLS) (Fig. 1c and Supplementary Fig. 1b,c) with a polydispersity index (PDI) of 0.12–0.25. Western blot experiments showed that expressions of exosome (CD9, CD63, TSG101) and microvesicle (ARF6) biomarkers in the CNP-treated group were significantly higher than in the untreated group, confirming the increase in secreted EVs (Supplementary Fig. 1d). Kinetic analyses further showed that voltage-optimized EV release peaked at 8 h after CNP induction, with continued secretion noted over the next 24 h (Supplementary Fig. 1e,f). Reverse transcription-quantitative polymerase chain reaction (RT–qPCR) showed that CNP-secreted EVs contained more than 200 times the COL1A1 mRNA than BEP-secreted EVs and 3,000-fold higher COL1A1 mRNA than EVs secreted from non-transfected cells (Fig. 1d). Bioanalyser assessment of gel agarose demonstrated full-length transcribed COL1A1 mRNA at ~4,000 nucleotides (Fig. 1e). EVs prepared by CNP exhibited structural stability when stored for preclinical administration at 4 °C with no changes in appearance, membrane and size properties when evaluated by cryo-electron microscopy (cryo-EM), atomic force microscopy (AFM) and NTA (Supplementary Fig. 2a–c). Furthermore, COL1A1 mRNA encapsulated within EVs was stable both at room temperature and at 4 °C, and also exhibited serum stability, thereby highlighting their potential for future clinical utility (Supplementary Fig. 2d,e).

Fig. 1: CNP generates large quantities of EVs loaded with COL1A1 mRNA.
figure 1

a, Schematic representation of CNP-generated EVs for targeted nucleic acid delivery. b, EV number per cell produced by untreated nHDFs in PBS buffer as control, BEP or CNP with PBS buffer only in 24 h (n = 3 for each group; *P = 0.048 Control vs BEP; **P = 0.005 Control vs CNP; ##P = 0.002 BEP vs CNP). c, Characterization of EVs by nanoparticle tracking analysis. d, RT–qPCR of COL1A1 mRNA reveals that EVs produced by CNP contain much larger quantities of transcribed mRNAs than EVs produced by BEP and Control groups in 24 h (n = 3 for all groups; **P = 0.002 Control vs BEP; ***P < 0.001 Control vs CNP; ###P < 0.001 BEP vs CNP). e, Gel electrophoresis of total RNA extracted from 3.78 × 108 EVs produced from CNP using COL1A1-GFP plasmid compared with synthetic in vitro-transcribed COL1A1 mRNA as control. f, Proliferation of nHDFs after treatment with different protein concentrations of COL1A1-EVs at 0 h, 24 h and 48 h (n = 4 for all groups; at 24 h: #P = 0.0107, 10 µg ml−1 vs 0 µg ml−1 of protein concentration of COL1A1-EVs after treatment; ##P = 0.002, 20 µg ml−1 vs 0 µg ml−1; ###P < 0.001, 30 µg ml−1 vs 0 µg ml−1; at 48 h: *P = 0.045, 10 µg ml−1 vs 0 µg ml−1 of COL1A1-EV treatment; ***P < 0.001, 20 µg ml−1 vs 0 µg ml−1; ***P < 0.001, 30 µg ml−1 vs 0 µg ml−1). All data are from three independent experiments and are presented as means ± s.e.m. Two-sided Student’s t-tests were used for the comparisons in b and d. One-way ANOVA was used for the comparisons in f. The schematic in a was created with

To assess the therapeutic potential of COL1A1 mRNA-containing EVs (COL1A1-EVs) in vitro, we treated cultured fibroblasts with COL1A1-EVs for 48 h (Supplementary Fig. 3a). Proliferation of fibroblasts was observed to increase with COL1A1-EV treatment in a dose-dependent manner (Fig. 1f). After treatment, elevated COL1A1 protein expression was observed in fibroblasts treated with COL1A1-EVs, as evidenced by increased co-localization of green fluorescent protein (GFP) with COL1A1 under immunofluorescence microscopy. In contrast, COL1A1 levels were substantially lower without co-localization of GFP in the non-treated group (Extended Data Fig. 1a,b). Successful mRNA delivery into recipient cells was further confirmed by elevated collagen mRNA expression and collagen protein levels measured via quantitative PCR and western blot analysis, respectively, in fibroblasts treated with COL1A1-EVs (Extended Data Fig. 1c,d and Supplementary Fig. 3b). Moreover, pro-collagen I, a precursor of COL1 protein, was significantly increased following COL1A1-EV treatment, indicating improved COL1 protein synthesis by the EV-treated fibroblasts compared with untreated fibroblasts (Extended Data Fig. 1e). Confocal microscopy imaging of cellular uptake of EVs indicated that EV cargo delivery into recipient cells is mediated by clathrin-mediated endocytosis (Supplementary Fig. 3c,d) and is capable of lysosome escape (Supplementary Fig. 4). Taken together, these findings demonstrate that functional COL1A1 mRNA can be stably encapsulated within EVs by CNP and that this COL1A1-EV delivery system can significantly increase COL1A1 protein expression in vitro.

In vivo kinetics of COL1A1 mRNA expression and protein translation following COL1A1-EV delivery

To understand the kinetics of EV-mediated COL1A1 mRNA delivery and protein formation in vivo, we delivered 2.7 × 109 copy number of COL1A1 mRNA COL1A1-EVs into the dermis of mice via an insulin needle syringe. The mice were killed over the next 14 d for histological analysis by RNAscope for COL1A1 mRNA quantification. COL1A1 mRNA was found to be significantly elevated in local skin tissue at 12 h after delivery, with notable decreases observed at 24 h and 48 h after injection, and a return to baseline COL1A1 mRNA levels by 96 h (Fig. 2a,b). In vivo translation of COL1A1 mRNA into protein was assessed by immunofluorescence microscopy of explanted tissue over a 30 d period. GFP+ COL1A1+ immunostained grafts were observed as early as 12 h after injection, with peak fluorescence at 4 d after delivery (Fig. 2c–e). GFP+ COL1A1+ protein grafts were observed to decrease in a time-dependent manner from day 4 to day 30 after injection as observed by immunofluorescence microscopy, with the majority of COL1A1-EV-derived protein turning over by day 30. These findings demonstrate that lower doses of COL1A1-EV delivery results in a 3–4 d peak of COL1A1 protein in recipient tissues, decreasing to baseline levels by day 30 after injection.

Fig. 2: In vivo kinetics of COL1A1-EV mRNA delivery and protein formation in murine skin.
figure 2

a, In situ hybridization of human COL1A1 mRNA by RNAscope after intradermal injection of COL1A1-EVs and saline control, measured at 12 h, 24 h, 48 h, 96 h, 7 d, 10 d and 14 d after injection. Scale bar,100 µm. b, Quantification of RNAscope results by average number of brown dots per cell. c, Immunofluorescence over time of COL1A1-EV-derived protein by visualization of co-localized GFP and COL1A1 protein (RFP). Scale bar, 100 µm. d, Fluorescence intensity quantification of COL1A1-GFP protein expression. e, Quantification of GFP+ cells confirms that the COL1A1-EV-derived protein grafts decrease in a time-dependent manner over 30 d. All data are from three independent experiments and are presented as means ± s.e.m.

In vivo therapeutic efficacy of COL1A1-EVs in a collagen-depletion skin-photoaging model

Dermal photoaging is characterized by the destruction of collagen and extracellular-matrix (ECM) proteins in the skin by sun exposure and UV irradiation, which causes the formation of wrinkles33,34. To assess the capacity of COL1A1 mRNA EVs to replace degraded collagen protein in vivo, we employed a previously described photoaging model in which athymic nude mice were treated with UV (311 nm) irradiation over an 8 week period, resulting in dermal wrinkles secondary to collagen depletion35 (Supplementary Fig. 5a). The efficacy of the 8 week UV irradiation course in degrading collagen and elastin fibres in murine skin was confirmed by histological and tissue analysis (Supplementary Fig. 5b–g). We next assessed the therapeutic potential of COL1A1-EVs to replace dermal collagen in UV-irradiated mice by treating the animals with a 5x injection course of either: (1) 2.7 × 109 copy number of COL1A1 mRNA encapsulated in EVs (COL1A1-EVs), (2) 2.7 × 109 copy number of COL1A1 mRNA encapsulated in lipid nanoparticles (COL1A1-LNPs), (3) control EVs without CNP cargo loading (unloaded EVs), (4) 0.05% retinoic acid topical treatment (RA) or (5) saline control. Wrinkle formation was tracked at days 0, 4, 7, 14, 21 and 28 after initiation of treatment, and all animals were killed at day 28 for histology and skin plaster replica analyses36 (Fig. 3a). Microscopic photography of dermal skin wrinkles over the treatment period demonstrated modest decreases in wrinkle number and area by day 28 for the COL1A1-LNPs, unloaded EVs and retinoic acid groups compared with the saline control group (Fig. 3b). In contrast, photoaged mice treated with COL1A1-EVs exhibited a reduction in wrinkle number and area beginning on day 7 after treatment initiation, with a significant reduction from day 14 onward to levels similar to those observed in unirradiated sham controls (Fig. 3c,d). Skin plaster replicas of the dorsal skin taken at day 28 after treatment initiation confirmed the effectiveness of COL1A1-EVs for treating photoaged skin as compared with saline control, retinoic acid, unloaded EV controls and COL1A1-LNPs (Extended Data Fig. 2). Noticeably, COL1A1-LNPs also exhibited a reduction in wrinkle number and area compared with saline control, retinoic acid and unloaded EV controls, although this effect was not as pronounced as with COL1A1-EVs. Skin treated with COL1A1-EVs and COL1A1-LNPs also demonstrated higher elasticity and firmness as measured by a cutometer and an Instron extensometer, respectively (Supplementary Fig. 6).

Fig. 3: COL1A1-EV mRNA delivery reduces dermal wrinkles in a UV-irradiation photoaging model.
figure 3

a, Schematic representation and timeline of 5 low-dose injections of COL1A1-EVs (2.7 × 109 copy number COL1A1 mRNA per dose). Skin tissue was collected at 28 d. b, Wrinkle formation was tracked at days 0, 4, 7, 14, 21 and 28 after initial delivery of COL1A1-EVs, COL1A1-LNPs, unloaded EVs, 0.05% retinoic acid (RA) or saline (n = 4 for all groups). The sham group comprised female nude mice that were not exposed to UV irradiation. Scale bar, 5 mm. c, Total numbers of dorsal-skin wrinkles quantified with custom software (n = 4 for all groups; ※※P = 0.0016 COL1A1-EVs vs Saline at day 7; ※※※P < 0.001 COL1A1-EVs vs Saline at days 14, 21 and 28; **P = 0.0089 COL1A1-LNPs vs Saline at day 21; **P = 0.0097 COL1A1-LNPs vs Saline at day 28; #P = 0.461 Unloaded EVs vs Saline at day 21; ##P = 0.0034 Unloaded EVs vs Saline at day 28; P = 0.0142 RA vs Saline at day 21; ††P = 0.0069 RA vs Saline at day 28). d, Quantification of wrinkle area on dorsal skin (n = 4 for all groups; ※※P = 0.0063 COL1A1-EVs vs Saline at day 7; ※※※P < 0.001 COL1A1-EVs vs Saline at days 14, 21 and 28; **P = 0.0082 COL1A1-LNPs vs Saline at day 21; ***P < 0.001 COL1A1-LNPs vs Saline at day 28; ††P = 0.0040 RA vs Saline at day 28; ##P = 0.0018 Unloaded EVs vs Saline at day 28). All data are from three independent experiments and are presented as means ± s.e.m. Two-way ANOVA was used for the comparisons in c and d. The schematic in a was created with

To assess the immunogenic side effects of COL1A1-LNPs and COL1A1-EVs, treated skin was excised from a subgroup of animals and analysed for inflammation 24 h after injection. Skin treated with COL1A1-LNPs exhibited redness and swelling as compared with sham controls and skin treated with COL1A1-EVs (Extended Data Fig. 3a). Flow cytometry and enzyme-linked immunoassay (ELISA) of tissue further revealed massive leucocytic infiltrate dominated by neutrophils in tissue that received COL1A1-LNPs (Extended Data Fig. 3b,c) and high levels of inflammatory cytokines such as TNF-α, IL-6, IL-1β and IFN-γ (Extended Data Fig. 3d–f). By comparison, tissue that received COL1A1-EV treatment did not exhibit a strong inflammatory reaction. These findings suggest that LNPs are substantially more immunogenic than EVs.

After assessment at day 28, a subset of animals was retained for 4 additional weeks to monitor wrinkle reduction duration. Dermal wrinkles were seen to reappear as early as 1 week later, beginning at day 35 after treatment initiation, and by day 56, dermal wrinkles were statistically indistinguishable from pre-treatment levels (Extended Data Fig. 4).

To evaluate dermal collagen engraftment from COL1A1-EV delivery, skin was excised from all groups at day 28 after treatment initiation and evaluated by immunofluorescence microscopy, as well as Masson trichrome staining. Histological analysis and fluorescence intensity quantification revealed that COL1A1 protein expression was significantly restored after COL1A1-EV treatment relative to other groups (Fig. 4a,b). Masson trichrome staining confirmed higher levels of collagen staining and dermal thickness (197.95 ± 24.46 μm COL1A1-EVs vs 95.19 ± 10.66 μm Saline at day 28, P < 0.001) in mice given COL1A1-EVs than in mice given saline control, retinoic acid, unloaded EVs or COL1A1-LNPs (Fig. 4c,d).

Fig. 4: Histologic analysis demonstrates collagen replacement in the dermis of photoaged mice treated with COL1A1-EV mRNA.
figure 4

a, Immunofluorescence staining for GFP and COL1A1 (RFP) protein in sham control group, saline control group, RA treatment group, unloaded EV treatment group, COL1A1-LNPs and COL1A1-EV treatment groups. COL1A1-EV-treated mice exhibited GFP+ COL1A1 protein grafts in the dermis and subcutis at 28 d after initiation of treatment. Scale bar, 200 µm. b, Fluorescence intensity of COL1A1 protein (RFP) for all treatment groups (n = 3 for all groups; ※※※P < 0.001 COL1A1-EVs vs Saline; ***P < 0.001 COL1A1-LNPs vs Saline; #P = 0.0386 Unloaded EVs vs Saline; P = 0.0218 RA vs Saline; +P = 0.0255 COL1A1-EVs vs COL1A1-LNPs). c, Representative Masson trichrome collagen staining of the epidermis, dermis and subcutis for all mouse groups. Collagen is blue. Scale bar, 200 µm. d, Quantification of dermal thickness showing increased collagen fibres in the COL1A1-EVs group (n = 3 for all groups; ※※※P < 0.001 COL1A1-EVs vs Saline; ***P < 0.001 COL1A1-LNPs vs Saline; #P = 0.0437 Unloaded EVs vs Saline; P = 0.0447 RA vs Saline; +++P < 0.001 COL1A1-EVs vs COL1A1-LNPs). All data are from three independent experiments and are presented as means ± s.e.m. One-way ANOVA was used for the comparisons in b and d.

Design of a microneedle-based system for the delivery of EV mRNAs

As wrinkles were found to return to baseline levels in photoaged animals within 1 month of COLA1-EV treatment cessation, we next aimed to improve the duration of protein replacement and the effectiveness of collagen engraftment by designing an HA microneedle formulation (COL1A1-EV MN) to improve EV-mediated mRNA tissue delivery. Custom microneedle patches were prepared by using a micromoulding method in which COL1A1-EVs were mixed with a 15% HA solution dissolved in PBS and cast into the tips of a polydimethylsiloxane (PDMS) mould by maintaining a vacuum for 30 min, after which 1 ml of 15 wt% HA solution was placed into the micromould and then transferred to 4 °C for 4 h for solidification (Fig. 5a). Each needle of the COL1A1-EV microneedle patch was moulded in a conical shape, with a circular diameter of 400 µm at their base and a height of 1,000 µm (Fig. 5b). The mechanical strength of patches with 10%, 15% or 20% HA loaded with EVs was evaluated with a tensile testing machine (Supplementary Fig. 7a). The load fracture force of our 15% COL1A1-EV MN was confirmed to be higher than the minimum average force needed for skin penetration (0.058 N)37 (Supplementary Fig. 7b). Haematoxylin and eosin (H&E) staining confirmed that the microneedles penetrated through the stratum corneum into the dermis (516 ± 76 μm)38 (Fig. 5c). COL1A1-EVs loaded into HA microneedles were found to have stable appearance and membrane integrity as evaluated by transmission electron microscopy (TEM) and AFM, as well as stable COL1A1 mRNA and particle size distribution (Supplementary Fig. 7c–f). For delivery into tissue, COL1A1-EV MN patches were pressed into the dorsal skin of mice, and the microneedle base was removed after 15 min (Supplementary Video 1). During this period, the microneedles dissolved completely, with no visible skin irritation or marking at the site of administration (Fig. 5d and Supplementary Video 2). To compare the tissue distribution of EVs delivered via microneedle with that of EVs delivered by insulin syringe, EVs were labelled with DiI (1,1′-dioctadecyl-3,3,3′,3′-tetramethylindocarbocyanine perchlorate; Beyotime, C1036)39 and administered intradermally under the dorsal skin of recipient mice, with subsequent histological assessment by immunofluorescence microscopy (Fig. 5e and Supplementary Fig. 8a,b). Tissue distribution analysis revealed that syringe needle injection resulted in uneven delivery of EVs with clumping in specific areas of the dermis and subcutis, whereas EVs delivered by microneedle were better dispersed in the dermis and subcutis (Fig. 5f and Supplementary Fig. 8c,d). Assessment of EV membrane integrity by cryo-EM showed less EV membrane lysis in the EVs delivered with the microneedle patch vs the syringe needle (18.1 ± 3.0% COL1A1-EV MN vs 28.3 ± 2.4% needle injection, P = 0.022) (Supplementary Fig. 8e,f). To assess whether the improved tissue distribution and EV membrane protection resulted in improved EV engraftment, we next employed serial in vivo fluorescence imaging of DiI-labelled EVs injected intradermally in mice over a 14 d period. Fluorescence imaging confirmed similar DiI EV signal between syringe-delivered EVs and microneedle-delivered EVs for the first 4 d after injection. However, the COL1A1-EV microneedle group had significantly higher fluorescence signal at day 4 (69.61 ± 6.4% COL1A1-EV MN vs 45.79 ± 2.5% needle injection, P < 0.001) and at day 7 (35.78 ± 5.9% COL1A1-EV MN vs 19.38 ± 3.8% needle injection, P < 0.001), lasting up to day 10 post injection (30.22 ± 2.6% COL1A1-EV MN vs 6.68 ± 1.3% needle injection, P < 0.001) (Fig. 5g,h). Ex vivo organ distribution analyses of DiI probe fluorescence over a 2 week time period revealed similar tissue distribution and signal decay across animals receiving intradermal injection and animals receiving microneedle delivery, suggesting that the reduction in signal intensity is due to metabolism of the DiI probe as shown in other similar studies40,41 (Supplementary Fig. 9a,b). Taken together, these findings indicate that use of the microneedle patch improves long-term retainment of EVs in tissue.

Fig. 5: A microneedle delivery system for improved EV distribution in tissue.
figure 5

a, Schematic illustration of microneedle fabrication. b, Microscope and scanning electron microscopy images of microneedle arrays. Scale bar, 500 µm. c, H&E-stained section of mouse skin shows penetration of single microneedle. Scale bar, 100 µm. d, Top: time course of HA EV microneedle tips pressed into skin; the microneedles dissolved within 15 min of application. Scale bars, 200 µm. Bottom: skin recovery after HA EV microneedle treatment shows minimal irritation. Scale bars, 5 mm. e, Skin histology of Dil-labelled EVs shows highly concentrated EVs (red) unevenly distributed in the subcutis after syringe needle injection, whereas microneedle-delivered EVs were more evenly distributed in tissue. Yellow dashed lines encircle representative subcutaneous hair bulbs and border representative rod-like portions of the follicle extending upward to the dermis. Scale bar, 100 µm. f, Representative EV distribution analysed by ImageJ software. g, In vivo fluorescence images of nude mice treated with intradermal injection or HA microneedle patch delivery of Dil-labelled EVs on days 1, 2, 4, 7, 10 and 14 after delivery (n = 3 for all groups). h, Quantification of fluorescence intensity over the 14 d treatment period. Data are normalized to the fluorescence intensity at day 1. All data are from three independent experiments and are presented as means ± s.e.m. ***P < 0.001 COL1A1-EV MN delivery group vs needle delivery group. Two-way ANOVA was used for the comparisons in h. The schematic in a was created with

Therapeutic efficacy of COL1A1-EV delivery via microneedles

We next tested whether use of our custom COL1A1-EV microneedle patch for delivering COL1A1-EVs could result in improved in vivo protein replacement in the skin of photoaged mice (Supplementary Fig. 10). Athymic mice were again subjected to 8 weeks UV irradiation and assigned to 1 of 5 treatment groups: (1) saline control, (2) COL1A1 mRNA EVs delivered by syringe needle, (3) HA microneedle control, (4) unloaded EVs delivered by HA microneedle (Unloaded EV microneedle) and (5) COL1A1 mRNA-loaded EVs delivered by HA microneedle (COL1A1-EV MN). All mice were given a single-dose injection of 22 × 109 copy number of COL1A1 mRNA (or an equivalent vehicle volume for control groups) at day 0 of the treatment timeline. After injection, all mice were monitored via microscopic photography of dorsal skin wrinkles for up to 3 months (Fig. 6a). Relative to syringe needle injection, which reduced wrinkle formation for up to 35 d before a return to pre-treatment baseline, delivery of COL1A1 mRNA by COL1A1-EV MN was found to substantially reduce wrinkle area and number for up to 70 d before a return to baseline levels (Fig. 6b,c). To further confirm these findings, a subset of mice from each group were killed at 1 month, 2 months and 3 months after treatment for skin replica plaster assessment of the dorsal skin and histology (Extended Data Fig. 5a–c). Both needle injection and COL1A1-EV MN treatments reduced wrinkle length and depth for up to 1 month post treatment (Extended Data Fig. 5d,e). However, only mice treated with COL1A1-EV MN showed significantly long-lasting reductions in wrinkle length and depth for up to the 2 month timepoint. To test whether collagen replacement and wrinkle treatment could be maintained for a longer period of time, we subjected additional animal cohorts to serial treatment of COL1A1-EVs delivered by syringe needle and microneedle every 30 d. Both COL1A1-EVs delivered by syringe needle and COL1A1-EV MN were found to reduce wrinkle number and area for as long as the animals received treatment, with COL1A1-EV MN animals demonstrating the most improvement (Extended Data Fig. 6).

Fig. 6: COL1A1-EV delivery via COL1A1-EV microneedle improves long-term treatment of photoaged skin.
figure 6

a, Long-term (91 d) observation of 4 treatment groups after a single injection: (1) saline control, (2) COL1A1-EVs (22 × 109 copy number COL1A1 mRNA) delivered by 28G syringe needle, (3) HA microneedle control, (4) unloaded EVs delivered by HA microneedle (Unloaded EV microneedle) and (5) COL1A1-EV MN (22 × 109 copy number COL1A1 mRNA), (n = 4 for all groups). Scale bar, 5 mm b, Quantification of total numbers of wrinkles (n = 4 for all groups; *P = 0.014 COL1A1-EV MN vs Saline at day 7; *P = 0.038 COL1A1-EV MN vs Saline at day 14; *P = 0.012 COL1A1-EV MN vs Saline at day 21; *P = 0.039 COL1A1-EV MN vs Saline at day 28; *P = 0.031 COL1A1-EV MN vs Saline at day 35; *P = 0.03 COL1A1-EV MN vs Saline at day 42; *P = 0.022 COL1A1-EV MN vs Saline at day 49; *P = 0.031 COL1A1-EV MN vs Saline at day 56; **P = 0.008 COL1A1-EV MN vs Saline at day 63; #P = 0.030 Needle injection vs Saline at day 7; #P = 0.041 Needle injection vs Saline at day 14; P = 0.047 HA microneedle vs Saline at day 14). c, Quantification of total dorsal skin wrinkle area (n = 4 for all groups; *P = 0.016 COL1A1-EV MN vs Saline at day 7; *P = 0.038 COL1A1-EV MN vs Saline at day 14; *P = 0.042 COL1A1-EV MN vs Saline at days 21 and 28; *P = 0.026 COL1A1-EV MN vs Saline at day 42; *P = 0.048 COL1A1-EV MN vs Saline at day 49; #P = 0.033 Needle injection vs Saline at day 7; #P = 0.031 Needle injection vs Saline at day 14; #P = 0.042 Needle injection vs Saline at day 21). All data are from three independent experiments and are presented as means ± s.e.m. Two-way ANOVA was used for the comparisons in b and c.

Histologic analysis of skin samples taken after treatment confirmed the lasting engraftment of GFP+ COL1A1 protein engraftment in the dermis and subcutis of mice in both the needle injection group and the COL1A1-EV microneedle group at 1 month after delivery (Fig. 7a). However, by the 2 month timepoint, only the mice in the COL1A1-EV microneedle cohort demonstrated GFP+ COL1A1 engraftment in the dermis and subcutis (Fig. 7b). At 3 months after delivery, skin samples in all groups failed to exhibit any GFP collagen engraftment (Fig. 7c). COL1A1+ GFP immunofluorescence intensity was quantified, supporting wrinkle microscopy findings which demonstrated effective treatment at days 30–60 and a return to pre-treatment baseline from days 70–90 (Fig. 7d). Immunohistochemical staining of COL1A1 protein and Masson trichrome staining of skin tissues confirmed that the amount of collagen in the dermis correlated with immunofluorescence microscopy findings and that among all cohorts, the COL1A1-EV MN group had the most abundant collagen fibres and highest dermal thickness (180.14 ± 21.46 μm COL1A1-EV MN vs 96.61 ± 14.00 μm Saline at 1 month, P < 0.001; 154.88 ± 8.27 μm COL1A1-EV MN vs 109.25 ± 10.86 μm Saline at 2 months, P < 0.05) (Fig. 7e,f). Taken together, these findings indicate that encapsulation and delivery of COL1A1 mRNA via our COL1A1-EV MN system can prolong collagen protein replacement in photoaged skin for more than twice the duration of syringe needle delivery.

Fig. 7: COL1A1-EV mRNA delivery via COL1A1-EV MN results in long-term protein replacement in skin.
figure 7

ac Immunofluorescence staining of GFP and COL1A1 (RFP) demonstrates GFP-positive COL1A1 protein grafts in the skin of mice receiving COL1A1-EVs via 28G needle injection and via COL1A1-EV MN for up to 1 month (28 d) after delivery. At 2 months (63 d), only mice that were treated with COL1A1-EV MN had long-term GFP-positive COL1A1 engraftment. No evidence of GFP-positive collagen protein could be detected in any mice by 3 months (91 d) after delivery. Scale bar, 200 µm. d, Quantification of GFP and COL1A1 (RFP) co-localized fluorescence signal demonstrates long-term COL1A1-EV-derived collagen engraftment in the skin of mice given COL1A1-EV MN vs mice given COL1A1-EVs via 28G needle injection and control groups (n = 3 for all groups; ***P < 0.001 COL1A1-EV MN vs Needle injection at 2 months). e, Immunohistochemical staining of COL1A1 protein and Masson trichrome staining at 1 month (28 d), 2 months (63 d) and 3 months (91 d). Scale bar, 200 µm. f, Quantification of dermal thickness by Masson trichrome staining (n = 3 for all groups; **P = 0.0062 COL1A1-EV MN vs Saline at 1 month; *P = 0.034 COL1A1-EV MN vs Saline at 2 months). All data are from three independent experiments and are presented as means ± s.e.m. Two-way ANOVA was used for the comparisons in d and f.


Extracellular vesicles have emerged as a next-generation drug-delivery system because of their inherent properties of biocompatibility, low immunogenicity and ability to be derived from healthy human cells12. Nonetheless, most studies focusing on nucleic-acid delivery so far have encapsulated small molecules in the 10–20 nt range such as microRNAs and small interfering RNA (siRNAs) as the payload, whereas larger nucleic acids such as mRNAs are seldom evaluated owing to the difficulty of loading them into EVs42. With the recent advancement in utility of mRNA-based therapeutics for treating human diseases, there has been increased interest in using EVs as an mRNA-delivery system. Recently, we developed an mRNA-loading technique that allows for the large-scale production of EVs containing intact endogenous mRNA for nucleic-acid therapy14. Here we have shown that mRNA-loaded EVs can be applied for protein-replacement therapy in a model of dermal collagen photodepletion17. We have shown that CNP is able to load high copy numbers of COL1A1 mRNA (~4,000+ nt) into EVs, which cannot be achieved by post-insertion loading methods43,44. In vivo results showed that these COL1A1-EVs can restore COL1A1 protein expression in the mouse skin after photoaging. We also examined the presence of COL1A1 mRNA and protein over time in vivo, and found that the corresponding protein was translated as early as 12 h after delivery with a peak at 4 d, and that this was sustained for several weeks, depending on dose. Several studies have generated ex vivo COL1A1 protein as a collagen filler; here we characterized the in vivo kinetics of exogenous COL1A1 mRNA delivery and protein expression in vivo45.

To adapt our approach for long-term protein replacement, we further developed a microneedle array for the delivery of COL1A1-EVs to allow for an even distribution of EVs into local tissue, with reduced membrane rupture. HA is a critical element of the extracellular matrix and has been verified to have excellent biocompatibility with skin tissues and various biomaterial systems46,47. By integrating EVs into an HA microneedle biomaterial, we were able to extend protein engraftment of COL1A1 to more than 60 d in the skin specimens evaluated. Notably, EVs not loaded with COL1A1 mRNA were also able to achieve a modest decrease in wrinkle number, which is consistent with studies finding that ‘empty’ EVs from certain parental cells are potential therapeutic candidates owing to endogenous cargos (still, the detailed mechanisms underlying such phenomena need further exploration)36.

Compared with the relatively long-lasting expression of DNA-based gene therapies, mRNA therapeutics may help advance gene therapy and lower the risks of adverse events owing to the retention of the mRNAs in the cytoplasm, without penetration into the nucleus48,49. Hence, mRNA-based modalities have advantages over DNA-based and virus-based gene therapies, such as the ability to bypass the traditional transcription process, and no risk of genomic integration (as opposed to the use of some adeno-associated virus vectors, which traditionally has been considered to have a low risk profile50). However, the clinical translation of mRNA therapeutics is still limited. For mRNA products developed using LNPs, the immunogenicity of PEGylation components on the LNP surface as well as some carrier formulations have been linked to inflammation and to numerous adverse safety events51,52. Indeed, when we compared the delivery of COL1A1-LNPs and COL1A1-EVs in vivo, we found that COL1A1-LNPs were able to yield collagen protein and reduce dermal wrinkles, but also caused notable inflammatory infiltrate in local tissue, whereas COL1A1-EVs did not. Several recent studies have shown that EVs are characterized by low levels of immunogenicity, with some preclinical reports and clinical trials ( NCT05191381, NCT05216562) suggesting that EVs derived from a variety of cell types have immunosuppressive effects, although it should be noted that these studies primarily used mesenchymal stromal cells and dendritic cells as donor cell types36,53,54.

Future challenges for the clinical application of COL1A1-EV MN include the optimization of microneedle geometry with more densely protected EVs as well as optimized storage conditions because mRNA is subject to rapid degradation when not retained at −80 °C or below55. Ideally, future EV systems can be packaged as ready-to-use aliquots and shipped at appropriate temperatures, which would greatly improve the practicability of COL1A1-EV MN systems for clinical use. We would also aim to expand therapeutic applications in our system to include items such as COL7 for genetic orphan diseases such as epidermolysis bullosa56. We have only delivered mRNA via the microneedle system, yet the system is also suitable for packaging other types of EV cargos, such as miRNAs and siRNAs, and other bioactive therapeutic agents such as peptides and proteins57,58,59. Although many challenges would need to be overcome before the microneedle-based EV-delivery system can be tried in humans, because of the improved biocompatibility and benign side-effect profile of EVs as compared with LNPs and adeno-associated virus (AAV), we believe that the system might constitute a universal nucleic-acid carrier for the treatment of a range of human diseases and conditions.


Cell culture

nHDFs (PCS-201-010) were purchased from ATCC and cultured in DMEM medium (Thermo Fisher) containing 10% heat-inactivated fetal bovine serum (FBS; 10099141C, Thermo Fisher) at 37 °C in humidified conditions equilibrated with 5% CO2.

Cellular nanoporation and electroporation

For CNP, a single layer of nHDF was seeded on a 1 cm × 1 cm 3D CNP surface for overnight incubation as previously described. Human COL1A1 cDNA (NM_000088.3) plasmid with GFP tag was purchased from Sino Biological (HG11776-ACG). Plasmids pre-loaded in PBS buffer were injected into individual cells via nanochannels using an electric field strength of 250 V cm−1 with 10 pulses at 10 ms per pulse, with a 0.1 s interval. Various electroporation conditions were tested to determine optimal conditions. BEP (gene pulser xcell, Bio-rad) was conducted using an electric field strength of 1,250 V cm−1 with 1 pulse of 20 ms. pCMV-COL1A1-GFP plasmids were prepared at a concentration of 500 ng ml−1 in PBS for transfection.

Collection and purification of EVs secreted by donor cells

Cells were cultured in DMEM medium containing serum. The cell culture medium containing serum was removed when conducting CNP. Cells were then washed with PBS 3 times and cultured in serum-free cell culture medium for 24 h after CNP. EVs were collected from cell culture supernatants. In brief, the cell culture medium (CCM) was centrifuged at 200 × g for 5 min to remove cells and debris, after which it was centrifuged again at 2,000 × g for 30 min. Amicon Ultra-4 centrifugal filter unit (10 kDa, Millipore, 801024) was used to concentrate the CCM. The EV sample was purified using total exosome isolation reagent (Invitrogen, 4478360). EV particle size and number were measured using NanoSight NS300 (Malvern). The RNA yield and size distribution were analysed using an Agilent 2100 Bioanalyzer with an RNA 6000 Pico kit (Agilent Technologies).

RT–qPCR of EV-containing RNA expression levels

The expression of human COL1A1 mRNA in EVs was measured using RT–qPCR following the manufacturer’s recommended protocol. In brief, total RNA from purified EVs was obtained using an RNA purification mini kit (Norgen Biotek, 55000) and a DNA removal kit (Norgen Biotek, 25720). A SuperScript III First-Strand Synthesis system (Invitrogen) was used to synthesize the first-strand complementary DNA, with random hexamers as primers. The expression of genes was measured using TB Green Premix Ex Taq II (Takara, RR820). The primer sequences used were as follows: COL1A1 (human), forward: 5′-CCTGGAAAGAATGGAGATGA-3′ and reverse: 5′-ACCATCCAAACCACTGAAAC-3′; Gapdh (human), forward: 5′-CAGCCTCAAGATCATCAGCA - 3′ and reverse: 5′ - AGAGGCAGGGATGATGTTCT - 3′ (Supplementary Table 1).

In situ hybridization for human COL1A1 mRNA in skin tissue

Nude female mice (10–12 weeks old) were anaesthetized with isoflurane, and the dorsal skin region was injected with 50 μl 2.7 × 109 copy number of CNP COL1A1 mRNA EVs. At predefined timepoints (12, 24, 48, 96 h, 7 d, 10 d, 14 d) following the skin injection, the skin was excised and put in 4% formaldehyde for fixation. Subsequently, the slices were embedded in paraffin further sectioned into 4 μm slices. RNAscope automated in situ hybridization assay for detection of human COL1A1 mRNA was performed using the HybEZ II Hybridization System (Advanced Cell Diagnostics (ACD)), and all in situ hybridization reagents were ACD products. In brief, target retrieval was performed at 95 °C for 15 min using Leica Epitope Retrieval Buffer 2, followed by protease treatment at 42 °C for 15 min. The probe (RNAscope Probe- Hs-COL1A1, 401891, ACD) was hybridized for 2 h at 40 °C followed by RNA scope amplification, and RNAscope 2.5 HD Assay BROWN kit was used for visualization of staining. RNAscope 2.5 LS probe-Rn-Ppib was used as negative control.

Immunofluorescence microscopy

Tissue sections were fixed with 4% paraformaldehyde for 20 min at room temperature and washed 3 times with PBS (Vetec) for 5 min each. Then the tissue was transferred to 0.2% Triton X-100 for 15 min (permeabilized), followed by blocking with BSA for 40 min and the addition of primary antibody (ab34710 and ab6556, Abcam) for overnight blocking at 4 °C. Finally, the secondary antibody (ab6939 and ab6717, Abcam) was added and the tissue sections placed at room temperature for 60 min. Tissue sections were then washed with PBS, added with DAPI (ThermoFisher) for nuclear staining and then mounted for observation.

Immunohistochemical staining and analysis

Tissue sections were fixed in 4% paraformaldehyde for 20 min, washed 3 times with PBS (pH 7.4) for 5 min each and then transferred to a retrieval box containing ethylenediaminetetraacetic acid (EDTA) (pH 9.0) antigen retrieval solution for antigen retrieval in a microwave oven. Subsequently, the sections were incubated in 3% hydrogen peroxide solution at room temperature for 25 min in the dark. After washing with PBS, the tissue was evenly covered with 3% BSA or 10% normal rabbit serum, blocked at room temperature for 30 min, added with primary antibody (ab34710 and ab6556, Abcam) and incubated overnight at 4 °C. Then, a secondary antibody (HRP-labelled) corresponding to the primary antibody was added to cover the tissue sections, which were then incubated at room temperature for 50 min. For 3,3'-diaminobenzidine (DAB) colour development, the slides were placed in PBS (pH 7.4) and washed 3 times with shaking on a de-staining shaker for 5 min each time. After the slices were slightly dried, the freshly prepared DAB colour developing solution was added dropwise in the circle. Developing time was controlled under the microscope. Haematoxylin, haematoxylin differentiation solution and hematoxylin blue-returning solution were sequentially added for nuclei counterstaining. Finally, the glass slides were placed in anhydrous ethanol and xylene for dehydration and sealing.


Animal work was approved by the Institutional Animal Care and Use Committee (IACUC) of Shenzhen Bay Laboratory (No. D2021-107) or partner laboratories. To create the skin-photoaging model, female nude mice (10–12 weeks old) were subjected to UVB irradiation of the dorsal skin every other day for 8 weeks as follows: mice were anaesthetized with 1.5% isoflurane, and then UV irradiation was delivered with a UV lamp (Philips; emission spectrum 311 nm) positioned 30 cm above the dorsal skin of the mice every other day for 8 weeks. The UV irradiation intensity, represented as the minimal erythemal dose (MED), was set at 1 MED during the first 2 weeks (60 mJ cm−2), elevated to 2 MED (120 mJ cm−2) in the third week, 3 MED (180 mJ cm−2) in the fourth week and 4 MED (240 mJ cm−2) during the fifth to eighth weeks of the experiment. The total irradiated UVB volume was approximately 80 MED. For the syringe-based treatment of photoaged skin, nude mice, after the 8 week irradiation period described above, were assigned to 1 of 6 treatment groups (4 mice each): (1) UVB irradiation + saline, (2) UVB irradiation + 0.05% retinoic acid, (3) UVB irradiation + unloaded nHDF-EVs delivered with a 32G Hamilton syringe, (4) UVB irradiation + 2.7 × 109 copy number COL1A1 mRNA COL1A1-LNPs, (5) UVB irradiation + 2.7 × 109 copy number COL1A1 mRNA COL1A1- EVs delivered with a 32G Hamilton syringe and (6) no UVB exposure (sham). Skin treatments took place on days 0, 4, 7, 14 and 21. The whole back skin was divided into three parts for analysis.

Skin replica

A SILFLO silicone replica and ring locator were purchased from Clinical & Derm. Replicas of the back (dorsal) skin of the mice were obtained at the end of the treatment period. The replicas were analysed by stereomicroscopy (Olympus SZX7), and corresponding images were analysed by ImageJ (NIH).

Preparation of COL1A1-EV MN patches and local tissue delivery

The fabrication of the microneedle patch was performed using a silicone micromould, with each needle cavity being 400 μm in a round base diameter and 1,000 μm in height. These needle cavities were arranged in a 10 × 10 array with 700 μm tip to tip spacing. For the preparation of the microneedle patch, 150 μl of 15 wt% HA solution was mixed with 50 μl EVs, kept under vacuum for 30 min and then transferred to 4 °C until it was deposited in the needle cavity. Finally, 1 ml of 15 wt% HA solution was loaded onto the micromould and the micromould placed in a drying chest with an attached fan to accelerate the evaporation process for solidification. After solidification, the microneedle patch was detached from the silicone mould for further use. To assess COL1A1-EV delivery via the customized microneedle patches, nude mice were irradiated for 8 weeks as described above and assigned to 1 of 5 treatment groups (4 mice each): (1) UVB irradiation + saline, (2) UVB irradiation + 22 × 109 copy number COL1A1 mRNA COL1A1-EVs delivered by 28G Hamilton syringe, (3) UVB irradiation + HA microneedle patch, (4) UVB irradiation + 1010 unloaded nHDF-EVs delivered via microneedle (unloaded EV microneedle) and (5) UVB irradiation + 15% HA microneedle mixed with 22 × 109 copy number COL1A1 mRNA COL1A1- EVs.

Fluorescence in vivo imaging

Microneedle-treated mice were imaged with an in vivo imaging system (IVIS Spectrum, Perkin Elmer) on days 1, 2, 4, 7, 10 and 14. The parameters were set as follows: exposure time 15 s, excitation 570 nm, emission 680 nm, 2F/stop and 13.6 cm field of view in the specified fluorescence imaging times. Quantitative analysis of RFP fluorescence intensity was performed by measuring the average radiation efficiency (photon s−1 cm−2 sr−1 µW−1) in a region of interest. Data were normalized to fluorescence intensity on day 1.

Statistical analysis

Quantitative data are represented as mean ± s.e.m. No data were excluded in this study. To analyse the statistical difference between two groups, two-sided Student’s t-tests were used for the comparisons. For two or more groups, one-way analysis of variance (ANOVA) with Tukey’s multiple-comparison test was used to analyse the statistical difference. Two-way ANOVA was chosen for analysing the statistical difference between data points in groups. P < 0.05 was considered statistically significant. GraphPad Prism 8.3 was used for data analysis.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.