Abstract

Background

   Osteoarthritis (OA) is a degenerative joint disease with an immense
   unmet medical need. FGF18 protein is a potential regenerative factor
   for cartilage repair. However, traditional protein delivery methods
   have limited efficacy due to the short lifetime and shallow
   infiltration.

Results

   In this work, we discovered that lipid nanoparticle (LNP) can
   infiltrate and deliver FGF18 mRNA deeper in the cartilage than
   proteins. After mRNA UTR optimization and chemical modification, the
   expression of FGF18 can last up to 6 days in the cartilage.
   Furthermore, delivering FGF18 mRNA activates FOXO3a-autophagy pathway,
   which protects against chondrocyte degeneration and senescence. Local
   intra-articular injection of FGF18 mRNA-LNP significantly alleviates OA
   symptoms in DMM and senile OA models. Sustained expression and
   accessibility of FGF18-mRNA to deeper chondrocytes makes LNP-mRNA more
   effective than FGF18 recombinant protein.

Conclusions

   In summary, this study presents a novel approach superior to
   recombinant protein alone and holds promise as a new therapeutic
   strategy for OA.

Graphical abstract

   [50]graphic file with name 12951_2025_3103_Figa_HTML.jpg

Supplementary Information

   The online version contains supplementary material available at
   10.1186/s12951-025-03103-9.

   Keywords: Osteoarthritis, FGF18, mRNA, Lipid nanoparticles, Autophagy

Introduction

   Osteoarthritis (OA) is a prevalent joint disease worldwide, affecting
   over 250 million people and imposing a significant burden on the
   economy and society [[51]1, [52]2]. Currently, most clinical treatments
   aim to alleviate pain symptoms in the early stages of the disease.
   Several biologic therapies [[53]3–[54]6] have shown promising efficacy
   in the osteoarthritis treatment. Despite encouraging preclinical
   results, a series of drugs failed in clinical development [[55]7].
   Numerous clinical deficiencies in drug efficacy can be attributed to
   insufficient drug delivery [[56]8].

   Rapid clearance and inadequate penetration depth are the main barriers
   for the effective drug delivery in cartilage. Even using
   intra-articular injections, the half-lives of many compounds are only
   2–4 h [[57]8]. The recombinant protein is also rapidly degraded, and
   only 7.7% of the recombinant protein remains on the surfaces of the
   femur and tibial cartilage after 24 h of injection [[58]9]. The short
   half-life reduces the function time of the therapy and restricts the
   overall efficacy. In addition, the majority of chondrocytes reside in
   the middle or deep zones of cartilage and covered by the negatively
   charged cartilage matrix, therefore, biological macromolecules are
   difficult to reach the chondrocytes [[59]10]. Thus, these unmet medical
   needs require novel strategies to improve the lifetime and penetration
   depth of recombinant proteins within the joint cavity.

   Recently, mRNA therapy has emerged as a promising approach [[60]11,
   [61]12]. In addition to the success of vaccines, mRNA has been applied
   in protein replacement therapy to rescue the protein deficiency, such
   as CoA carboxylase, Methylmalonyl-coenzyme A mutase (MUT), ornithine
   transcarbamylase (OTC) [[62]13]. The mRNA does not integrate into the
   genome. Once it enters the cytoplasm, it can be directly translated
   into proteins. Therefore, mRNA-based therapeutics have shown good
   safety profiles [[63]11]. However, negative charge and instability of
   mRNA, coupled with its susceptibility to degradation by nucleases and
   immunogenicity, highlight importance of a suitable delivery system
   [[64]14–[65]17]. Lipid nanoparticles (LNPs) can encapsulate the mRNA
   and neutralize the negative charge to extend its life in the patient
   [[66]18]. LNP are usually composed of four lipids, i.e. cationic
   ionizable lipids, phospholipids, PEG lipids, and cholesterol, which are
   widely used as an efficient delivery system for mRNA [[67]11,
   [68]19–[69]21]. The ionizable cationic lipids significantly extends the
   half-life of mRNA, protecting mRNA from nuclease-mediated degradation.
   In body fluids with a neutral pH, LNPs do not have any charge and
   therefore, do not interact with negatively charged red blood cells.
   However, after they are taken up by cells, LNPs become positively
   charged inside the acidic endosomal environment, helping mRNA to escape
   from the endosomes. In addition, modifying antibodies on the surface of
   LNP can also enhance the targeted delivery of mRNA [[70]22]. Overall,
   using LNP for mRNA delivery can overcome physiological barriers and
   enable the targeted delivery and translation of mRNA at the site of
   interest [[71]11]. While LNPs have been extensively studied in soft
   tissues [[72]19, [73]23–[74]25]. delivering drugs to cartilage is a
   challenging task due to its unique inorganic and organic composition.

   Recent studies have revealed that FGF18 plays an important role in the
   development and progression of osteoarthritis [[75]26–[76]29]. The FGF
   family and its receptors, known as FGF receptors (FGFRs), aid in the
   differentiation of chondrocytes and protect articular cartilage from
   degeneration during bone and cartilage development [[77]30, [78]31]. A
   multi-cohort genome wide association meta-analysis of osteoarthritis
   reported that single nucleotide polymorphism (SNP) of FGF18 gene
   positively correlate with osteoarthritis [[79]32]. Studies on FGFR3
   conditional knockout mice in chondrocytes showed increased severity of
   osteoarthritis after DMM (destabilization of medial meniscus) modeling
   [[80]33–[81]35]. Recombinant FGF18 protein has shown clinical potential
   for the treatment of osteoarthritis [[82]3, [83]36–[84]39].

   OA is prevalent among the elderly, and cellular senescence is a
   significant characteristic of OA chondrocytes [[85]40, [86]41].
   Autophagy, a normal cellular metabolic process, mediates the travelling
   of cellular components to lysosomes, promoting cell survival under
   stress conditions. Impaired autophagy is a key contributor to aging
   [[87]42]. In comparison to young mice, the expression of
   autophagy-related genes is significantly reduced in the cartilage of
   senile mice [[88]43]. These changes are associated with increased
   levels of apoptosis and cartilage degeneration. Similar findings have
   been observed in post-traumatic OA mouse models [[89]44], suggesting
   the importance of autophagy in both senescence and OA. Therefore,
   targeting chondrocyte autophagy is a promising direction for rescuing
   cartilage from senescence.

   In this study, we are motivated to investigate the efficacy of LNP
   encapsulated FGF18 mRNA in the OA treatment. Interestingly, we found
   that the LNP was able to penetrate deeper into the cartilage than FGF18
   protein alone, resulting in better delivery to the joint cavity. By
   optimizing the UTR region and chemical modification, we were able to
   increase FGF18 expression by 10-fold compared to the intrinsic UTR of
   FGF18. LNP-FGF18 mRNA showed protective effects against chondrocyte
   degeneration and senescence in different models, including inflammatory
   chondrocytes and two mouse OA models. In summary, our study presents a
   novel approach to effectively treat chondrocyte senescence and
   degeneration in the cartilage by delivering LNP-encapsulated FGF18 mRNA
   in vitro and in vivo.

Results

FGF18 expression level decreases in cartilage inflammation and senescence
microenvironment

   We conducted a comprehensive investigation into the expression levels
   of FGF18 in various osteoarthritis models. Our study included cellular
   cultures, murine models, and clinical specimens. Initially, we
   performed RNA sequencing on primary chondrocytes in an inflammatory
   environment. The decreased expression of cartilage matrix
   synthesis-related genes (SOX9, COL2A1 and ACAN) and the increased
   expression of matrix degradation-related genes (MMPs) confirm the
   successful stimulation by inflammatory factors. Results showed a 75%
   reduction in the expression level of FGF18 (Fig. [90]1A). We also
   stimulated ATDC5 chondrocytes with varying concentrations of IL-1β and
   TNF-α for 24 h. Results revealed an 80% decrease in mRNA expression of
   SOX9 and COL2A1, a 10-fold increase in MMP13 expression, and
   concomitant 60% reductions in FGF18 expression level (Fig. [91]1B). In
   the H[2]O[2]-induced cellular senescence model, the expression level of
   FGF18 decreased by 60% (Supplementary Fig. [92]1A). Results from the
   GSE database also indicate that the expression level of FGF18 in
   chondrocytes of elderly individuals is only 25% of that in young
   individuals (Supplementary Fig. [93]1B), confirming the downregulation
   of FGF18 in senescent microenvironment. To investigate the role of
   FGF18 in osteoarthritis, we established two distinct mouse models:
   destabilization of the medial meniscus (DMM) and a naturally aging
   model. Histological examinations (Safranine O/Fast Green) validated the
   successful creation of the arthritis models, with apparent cartilage
   abrasion in DMM and aging mice. (Fig. [94]1C-D, F-G).
   Immunohistochemical analyses revealed a 50% decrease in FGF18 positive
   cell proportion in both inflamed and senescent cartilage (Fig. [95]1C,
   E, F, H). Finally, we collected tibial plateau samples from patients
   undergoing total knee arthroplasty, which revealed a 50% decrease in
   FGF18^(+) positive cell proportion within damaged osteoarthritic
   cartilage (Fig. [96]1I-K). These results substantiate the proposition
   that the FGF18 experiences diminished expression within the
   microenvironment of joint cartilage inflammation and senescence while
   FGF18 supplementation could promote cartilage repair and delay the
   progression of OA.

Fig. 1.

   [97]Fig. 1
   [98]Open in a new tab

   FGF18 expression level decreases in inflammatory and senescent
   microenvironments. (A) Heatmap of relevant differential expression
   genes (DEGs) in chondrocyte RNA sequencing after TNF-α treatment. (B)
   mRNA expression levels of SOX9, COL2A1, MMP13, and FGF18 after
   stimulation with different concentrations of inflammatory factors
   (n = 3). (C-E) Safranin O/Fast Green staining, and FGF18
   immunohistochemical staining and corresponding quantitative analysis in
   knee joints of DMM model (n = 6). Scale bar: 200 μm. (F-H) Safranin
   O/Fast Green staining, immunohistochemical staining of FGF18 and
   quantitative analysis in knee joint cartilage of senile model (n = 6).
   Scale bar: 200 μm. (I-K) Safranin O/Fast Green staining,
   immunohistochemical staining of FGF18, and quantitative analysis of
   clinical specimens (n = 6). Scale bar: 500 μm and 250 μm. Data are
   shown as mean ± SD. P value was calculated by one-way ANOVA, followed
   by Tukey’s post hoc tests or Student’s t test. *P < 0.05, **P < 0.01,
   ***P < 0.001, and ****P < 0.0001

LNP-FGF18 mRNA delivers FGF18 to the deep zone of cartilage

   Penetration depth is a key factor the successful delivery of
   chondrocyte repair. We are wondering about the penetration depth of LNP
   and FGF18 proteins. We encapsulated the FGF18 mRNA within lipid
   nanoparticles (LNPs) (Fig. [99]2A). The prepared LNPs have an average
   size of 128 nm with a polydispersity index (PDI) of 0.019, showing a
   uniform particle size distribution (Fig. [100]2B). The Zeta potential
   is -4.03 ± 1.89 mV under physiological pH 7.4 conditions, and this
   value shifts to 22.0 ± 4.76 mV at pH 5.0, which is the pH value in
   endosomes. The regular spherical shape of LNPs was analyzed using
   cryo-transmission electron microscopy (Fig. [101]2C). Importantly, cell
   viability staining and CCK8 assays provided compelling evidence that
   8 μg/ml of LNPs were non-toxic to the chondrogenic cell line, i.e.
   ATDC5 cell and 293T cells (Supplementary Fig. [102]2). Furthermore, we
   labeled FGF18 with Cy3 fluorescence and LNP with Dil, respectively, to
   trace their infiltration depth in the cartilage. Interestingly,
   recombinant FGF18 protein remains on the surface of the cartilage
   layer, whereas LNP-mRNA can penetrate five times deeper into the
   cartilage up to nearly 50 μm in both young and old mice cartilage
   (Fig. [103]2D, E). Furthermore, collagen fibril network pore size of
   young and old mice was observed and measured with SEM (Fig. [104]2F).
   Although collagen network in aged mice is denser and has smaller pore
   sizes compared to young mice (ranging from 102 nm to 260 nm), it
   remains similar in diameter to prepared LNPs. These results suggest
   that the LNP-mRNA system possesses superior penetration capabilities
   which enable the LNP to act on deeper chondrocytes.

Fig. 2.

   [105]Fig. 2
   [106]Open in a new tab

   LNP-mRNA penetrates into cartilage and expresses the target protein.
   (A) Flowchart of the LNP-mRNA System Preparation. (B) Particle size of
   LNP. (C) Representative cryo-transmission electron microscopy images of
   LNP, Scale bar: 100 nm. (D, E) Representative immunofluorescence images
   of the penetration depth of Cy3-labelled FGF18 and Dil-labelled LNP in
   the cartilage of mice of different ages and quantitative analysis
   (n = 4). White dashed line represents the boundary between the
   cartilage and the subchondral bone. Scale bar: 10 μm. (F)
   Representative SEM images of collagen fiber networks in the cartilage
   of mice of different ages. Scale bar: 1 μm. (G-H) Luminescence
   intensity at different time points after transfection with lipo3000 and
   LNP systems in 293T and ATDC5 cells (n = 4). (I) Representative
   fluorescent images of ATDC5 cells transfected with varying doses of
   EGFP mRNA-LNP at different time points. Scale bar: 50 μm. (J-K)
   Bioluminescence imaging and quantitative analysis of signal flux at
   different time points after intra-articular injection of LNP-Fluc mRNA
   in mouse knee joints (n = 3). Data are shown as mean ± SD. P value was
   calculated by one-way ANOVA, followed by Tukey’s post hoc tests.
   *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001

LNP-FGF18 mRNA enables considerable expression of FGF18 in joint cavity

   We further compared the transfection efficiency between LNPs and
   Lipofectamine 3000 in the chondrogenic cell (ATDC5) and 293T cells. We
   used different concentrations of Fluc mRNA and our findings revealed
   that the LNP system had a superior transfection efficiency in both cell
   types, especially in ATDC5 cells, when compared to Lipofectamine 3000
   (Fig. [107]2G, H). Additionally, we observed that mRNA signals from LNP
   transfection at the 72-hour time point were approximately half of what
   was observed at the 24-hour point, whereas Lipofectamine 3000-mediated
   transfection showed a more significant decline at the 72-hour time
   point. Representative fluorescence microscopy images were applied to
   evaluate the transfection of 293T cells and ATDC5 cells with EGFP
   mRNA-LNP at varying doses and time points. The fluorescence results
   align with the signal trends observed for Fluc mRNA (Fig. [108]2I and
   Supplementary Fig. [109]3). These results suggest that the LNP-mRNA
   system is highly effective and has a long-lasting release capability in
   cells.

   Subsequently, our investigation explored the duration of the LNP mRNA
   system’s expression in a murine model. We administered 3 μg of LNP-Fluc
   mRNA into the knee joint of mice and monitored Fluc expression at
   various intervals using bioluminescence imaging. Our findings confirmed
   that the LNP mRNA system can sustain localized knee joint expression
   for up to 6 days (Fig. [110]2J, K). Monitoring bioluminescence in other
   organs of the mice revealed minimal evidence of LNP metabolism through
   the liver. The signal diminished after 48 h, indicating no side effects
   of LNP on other organs (Supplementary Fig. [111]4). On the other hand,
   injecting mRNA directly into the knee joint resulted in a significantly
   reduced Fluc signal. The signal persisted for only 48 h (Supplementary
   Fig. [112]5). These results collectively validate the efficacy and
   protracted release capability of the LNP-mRNA system.

   To enhance the expression and secretion of FGF18 protein, we optimized
   the 5’ untranslated region (UTR) of FGF18 mRNA. Plasmids containing
   varying UTRs, including mouse UTR (mUTR), synthetic UTR (SynUTR) and
   human UTR (hUTR), were transfected into ATDC5 and 293T cells,
   respectively. RT-qPCR analysis definitively demonstrated 400-fold
   increase in FGF18 plasmids (Fig. [113]3A). SynUTR, in particular, had
   the most significant impact on enhancing FGF18 protein expression,
   resulting in a 1000-fold increase (Fig. [114]3B-C, Supplementary
   Fig. [115]7A, B). Dot blot experiments on cell culture supernatants
   also confirmed that synUTR plasmid transfection resulted in the highest
   concentration of secreted FGF18. regardless of whether it was in ATDC5
   or 293T cells (Fig. [116]3D, Supplementary Fig. [117]6A, Supplementary
   Fig. [118]7C, D). We then encapsulated UTR-modified FGF18 mRNA within
   LNPs and measured the particle size, zeta potential, and encapsulation
   efficiency of LNPs containing different UTRs. The results indicate that
   the particle sizes of LNPs with various UTRs were all within the range
   of 120–130 nm, with zeta potentials between − 10 mV and 0 mV, and
   encapsulation efficiencies exceeding 90%. These findings suggest that
   while different UTRs of FGF18 mRNA can have some impact on LNP
   characterization, the effects are relatively minor. Transfection
   experiments in ATDC5 cells were then conducted at different time
   points. Notably, the results showed that FGF18 expression and secretion
   in the culture medium were positively correlated with the concentration
   of LNPs and negatively correlated with transfection time
   (Fig. [119]3E-K, Supplementary [120]6B, C). Specifically, 3 μg/mL
   LNP-FGF18 mRNA would generate 100 ng/ml FGF18 after 24 h. This specific
   concentration of LNP-FGF18 mRNA with SynUTR was subsequently utilized
   in all subsequent experiments.

Fig. 3.

   [121]Fig. 3
   [122]Open in a new tab

   Optimization of UTR regions of FGF18 mRNA to obtain sustained secretion
   of FGF18. (A-C) Detection and quantitative analysis of FGF18 mRNA and
   protein after transfection with plasmids containing different UTRs
   (n = 3). (D) FGF18 secretion was quantified in the cell culture
   supernatant of cells, which were tranfected with plasmids containing
   different UTRs (n = 3). (E-G) Detection and quantitative analysis of
   FGF18 mRNA and protein after transfection with different concentrations
   of LNP-FGF18 mRNA (n = 3). (H-I) Time course of FGF18 expression after
   transfection with LNP-FGF18 mRNA and quantitative analysis (n = 3). (J,
   K) Measurement of FGF18 content in the cell culture supernatant after
   transfection with different concentrations or durations of LNP-FGF18
   mRNA (n = 3). Data are shown as mean ± SD. P value was calculated by
   one-way ANOVA, followed by Tukey’s post hoc tests. *P < 0.05,
   **P < 0.01, ***P < 0.001, and ****P < 0.0001

LNP-FGF18 mRNA promotes the proliferation of chondrocytes by rescuing them
from degeneration and senescence

   FGF18 protein is known for its ability to mitigate cartilage
   extracellular matrix degradation and stimulating matrix synthesis
   [[123]36], but the effects in the LNP-mRNA format is unclear. Motivated
   by the superior infiltration depth and duration, we compared the
   effects of recombinant FGF18 protein and LNP-FGF18 mRNA in cells
   cultured in inflammatory (10ng/mL IL-1β) or non-inflammatory
   conditions. Both of them enhanced the mRNA expression and protein
   translation of genes associated with cartilage matrix synthesis, among
   which COL2A1 and SOX9 increased by up to 2-fold, while simultaneously
   suppressed the mRNA linked to matrix degradation and fibrosis, such as
   MMP3, MMP13, and COL1A1, by over 50% compared to solely IL-1β group
   (Fig. [124]4A-C). Meanwhile, immunofluorescence results show a 30%
   increase in SOX9 positive cell proportion and COL2A1 fluorescent
   intensity after FGF18 or LNP-FGF18 mRNA treatment (Fig. [125]4D-F).
   Micromass culture experiments consistently demonstrated that both
   LNP-mRNA and FGF18 protein effectively augmented the secretion of
   cartilage cell matrix and rescued matrix depletion under inflammatory
   conditions after 4, 6, and 8 days (Fig. [126]4G-J). IOD value of
   toluidine blue staining increases by 20% after FGF18 or LNP-FGF18 mRNA
   treatment (Fig. [127]4G, H) and that of alcian blue also elevates
   20–50% depending on different duration compared to control group or
   solely IL-1β group (Fig. [128]4I, J). Our findings suggest that the LNP
   mRNA exhibited similar effects to those of FGF18 recombinant protein in
   cells.

Fig. 4.

   [129]Fig. 4
   [130]Open in a new tab

   Recombinant FGF18 protein and LNP-FGF18 mRNA alleviate chondrocyte
   inflammation-induced degeneration. (A-C) mRNA transcription and protein
   expression levels of COL2A1, SOX9, MMP3, MMP13, and COL1A1 after
   treatment with inflammatory factors, FGF18 or LNP-FGF18 mRNA (n = 3 or
   4). (D-F) Immunofluorescence and quantitative analysis of COL2A1 and
   SOX9 expression levels in ATDC5 chondrocytes treated with FGF18 or
   LNP-FGF18 mRNA (n = 4). Scale bar: 50 μm. (G-J) Alcian Blue and
   Toluidine Blue staining of ATDC5 chondrocyte micromass cultures at
   different time points treated with FGF18 or LNP-FGF18 mRNA and
   quantitative analysis (n = 4). Data are shown as mean ± SD. P value was
   calculated by one-way ANOVA, followed by Tukey’s post hoc tests.
   *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001

   To comprehensively explore the mechanism by which LNP-FGF18 mRNA
   affects chondrocytes and its role in preventing cartilage degeneration,
   we applied RNA sequencing to analyze ATDC5 chondrocytes treated with
   LNP-FGF18 mRNA for 24 h (Supplementary Fig. [131]8). The genes related
   to cell cycle and senescence such as E2F1 and CCND1 were enriched
   (Fig. [132]5A-C). We then conducted further senescence-related studies
   on primary murine chondrocytes. RT-qPCR and western blot experiments
   revealed that LNP and FGF18 could downregulate the elevated expression
   of P16^INK4a, P21, and P53 by 40–60% compared to solely IL-1β group
   (Fig. [133]5D-F). In line with these results, SA-β-gal staining
   experiments showed that treatment with FGF18 mRNA and FGF18 protein
   under inflammatory conditions reduced the number of senescent primary
   murine chondrocytes by 50%. Moreover, EdU assays and Ki67
   immunofluorescence experiments conclusively demonstrated that both
   FGF18 and LNP-FGF18 mRNA could enhance the proliferation potential of
   primary chondrocytes, with proportions of EdU-positive and
   Ki67-positive cell proportions increased by approximately 100% compared
   to the solely IL-1β group (Fig. [134]5G-J).

Fig. 5.

   [135]Fig. 5
   [136]Open in a new tab

   Recombinant FGF18 protein and LNP-FGF18 mRNA inhibit chondrocyte
   senescence induced by inflammatory stimulation. (A, B) KEGG pathway
   enrichment analysis and GSEA(Gene Set Enrichment Analysis) of
   transcriptome sequencing data of ATDC5 chondrocytes stimulated with
   LNP-FGF18 mRNA. (C-E) mRNA and protein expression levels, along with
   quantitative analysis, of senescence-related genes in chondrocytes
   treated with FGF18 or LNP-FGF18 mRNA (n = 3 or 4). (F-I)
   Immunofluorescence staining for Ki67, EdU assays, and
   senescence-associated β-galactosidase staining after stimulation with
   FGF18 or LNP-FGF18 mRNA, along with quantitative analysis (n = 4).
   Scale bar: 50 μm and 200 μm. Data are shown as mean ± SD. P value was
   calculated by one-way ANOVA, followed by Tukey’s post hoc tests.
   *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001

   Afterwards, we attempted to understand the molecular mechanisms behind
   the recovery of cartilage senescence and degeneration with the help of
   FGF18. The RNA sequencing data showed that there was a significant
   increase in autophagy and FOXO signaling pathway after FGF18 LNP
   treatment (Fig. [137]5A). Notably, FGF18 is recognized as a gene
   associated with longevity and autophagy [[138]45], prompting us to
   explore into the expression patterns of autophagy-related genes. Our
   findings indicated that the addition of FGF18 mRNA led to an increase
   in FOXO3a and BECLIN1 expression, which in turn increased the LC3B II/I
   ratio (Fig. [139]6A, B). We also observed a decrease in the number of
   and autolysosomes in an inflammatory environment through
   immunofluorescence and transmission electron microscopy (Fig. [140]6C).
   Importantly, the presence of FGF18 protein and FGF18 mRNA significantly
   improved these changes.

Fig. 6.

   [141]Fig. 6
   [142]Open in a new tab

   Recombinant FGF18 and LNP-FGF18 mRNA activate chondrocyte autophagy
   pathway to counteract aging and degeneration. (A, B) Protein expression
   and quantitative analysis of FOXO3a and autophagy-related genes in
   ATDC5 chondrocytes after stimulation with FGF18 or LNP-FGF18 mRNA
   (n = 3). (C) Immunofluorescence staining of LC3 and representative
   images of autophagosomes by transmission electron microscopy after
   stimulation with FGF18 or LNP-FGF18 mRNA. Scale bar: 10 μm, 500 nm. (D,
   E) Protein expression and quantitative analysis of FOXO3a in FOXO3a
   knockdown cells (n = 3). (F, G) Expression and quantitative analysis of
   BECLIN1, LC3B, P16, SOX9 and COL2A1 in ATDC5 chondrocytes under the
   influence of IL-1β (10ng/mL), FOXO3a knockdown, 3-MA (5mM), and
   LNP-FGF18 mRNA (n = 3). (H-J) Alcian blue and toluidine blue staining
   for collagen in chondrocytes along with quantitative analysis (n = 4).
   (K-L) SA-β-gal staining, LC3B immunofluorescence, transmission electron
   microscopy representative images and related quantitative analysis
   (n = 4). Scale bar: 200 μm, 10 μm, 500 nm as shown in the figures. Data
   are shown as mean ± SD. P value was calculated by one-way ANOVA,
   followed by Tukey’s post hoc tests. *P < 0.05, **P < 0.01,
   ***P < 0.001, and ****P < 0.0001

   To elucidate the relationship between autophagy, chondrocyte
   senescence, and degeneration, we constructed FOXO3a knockdown cell
   lines (Fig. [143]6D-E) and imposed autophagy inhibition using 3-MA as
   inhibitor [[144]46]. We used a range of techniques, including western
   blot, micromass culture, LC3B immunofluorescence, SA-β-gal staining,
   and transmission electron microscopy to examine the effects of FGF18
   and LNP treatment on chondrocyte autophagy. we found that FGF18 mRNA
   offered protection to chondrocytes from degeneration and senescence by
   enhancing autophagy. The expression levels of cartilage matrix
   synthesis-related genes COL2A1 and SOX9 increased by 100%. The
   expression of BECLIN1 and LC3B II/I ratio was restored, while the
   expression of the senescence gene P16 decreased by 50%. The proportion
   of senescent cells decreased by 60%, significantly enhancing the matrix
   secretion function and autophagosome number in chondrocytes. When we
   applied FOXO3a knockdown and 3-MA treatment to these cells, the
   protective effects of FGF18 mRNA were reversed under inflammatory
   conditions, as seen through SA-β-gal staining and micromass culture
   (Fig. [145]6F-J). These results suggest that FGF18 mRNA protects
   cartilage cells from degeneration and senescence by enhancing
   autophagy.

LNP-FGF18 mRNA alleviates the progression of osteoarthritis in both DMM and
senile osteoarthritis models

   To further evaluate the efficacy of LNP-FGF18 mRNA in vivo, we
   established two distinct osteoarthritis models, i.e. natural senility
   and DMM. Weekly intra-articular injections of 500ng FGF18 protein or
   6 μg LNP-FGF18 mRNA were administered to the knee joint cavity
   (Fig. [146]7A). Biosafety experiments confirmed that there was no side
   effect in other organs after injection (Supplementary Fig. [147]9). Von
   Frey and hotplate experiments confirmed that both FGF18 and LNP-FGF18
   mRNA significantly reduced lower limb pain hypersensitivity induced by
   DMM (Fig. [148]7B, C). However, the LNP-mRNA group has significantly
   higher Von Frey threshold and response time in hotplate compared to the
   protein group. Gait analyses revealed a 20% improvement in parameters
   such as max contact area, print area, mean intensity and duty cycle in
   mice treated with FGF18 protein or LNP-FGF18 mRNA compared to the DMM
   or senile groups (Figs. [149]7D and E and [150]8A and B).

Fig. 7.

   [151]Fig. 7
   [152]Open in a new tab

   Rescue of osteoarthritis induced by DMM using FGF18 and LNP-FGF18 mRNA.
   (A) Experimental flowchart of the animal study. (B, C) Quantitative
   analysis of Von Frey and hotplate experiments at different time points
   (n = 15 or 18). (D, E) Representative gait trajectory and quantitative
   analysis of max contact area, print area, mean intensity, and duty
   cycle in different mice groups. (F) Representative images of X-ray,
   micro-CT reconstruction and subchondral bone reconstruction, synovial
   HE staining, cartilage Safranine O/Fast Green staining, and
   immunohistochemical staining for Ki67, LC3 and FGF18. Scale bar:
   100 μm. (G) Quantitative analysis of subchondral bone BV/TV, trabecular
   number, thickness, and separation (n = 6 or 7). (H) Quantitative
   analysis of osteophyte number (n = 6 or 7). (I-L) Quantitative analysis
   of synovial scores, OARSI scores, cartilage thickness, and area (n = 6
   or 7). (M-O) Quantitative analysis of FGF18 (M), Ki67 (N), LC3 (O)
   positive cells (n = 6 or 7). Data are presented as mean ± SD. P value
   was calculated by one-way ANOVA, followed by Tukey’s post hoc tests,
   Kruskal-Wallis test and Mann-Whitney test. *P < 0.05, **P < 0.01,
   ***P < 0.001, and ****P < 0.0001

Fig. 8.

   [153]Fig. 8
   [154]Open in a new tab

   Rescue of osteoarthritis symptoms in senile mice by FGF18 and LNP-FGF18
   mRNA. (A, B) Representative gait trajectory and quantitative analysis
   of max contact area, print area, mean intensity, and duty cycle in
   different mice groups (n = 15 or 18). (C) Representative images of
   X-ray, synovial HE staining, cartilage Safranin O/Fast Green staining,
   and immunohistochemical staining for Ki67, LC3, and FGF18. Scale bar:
   100 μm. (D-G) Quantitative analysis of synovial scores (D), OARSI
   scores (E), cartilage thickness (F), and area (G). (H-J) Quantitative
   analysis of FGF18 (H), Ki67 (I), LC3 (J) positive cells (n = 4 or 6).
   Data are presented as mean ± SD. P value was calculated by one-way
   ANOVA, followed by Tukey’s post hoc tests, Kruskal-Wallis test or
   Mann-Whitney test. *P < 0.05, **P < 0.01, ***P < 0.001, and
   ****P < 0.0001

   Radiological results indicate that intra-articular injections of FGF18
   protein and LNP-FGF18 mRNA could reduce subchondral bone sclerosis
   caused by the DMM model. This treatment results in a decrease in BV/TV
   and bone trabecular thickness compared to the DMM group. However, the
   trabecular number remained unchanged after FGF18 or LNP administration
   (Fig. [155]7F, G). Additionally, the number of osteophytes was reduced
   by 60% reduction after LNP-FGF18 mRNA injection, superior to 30%
   reduction after FGF18 protein injection (Fig. [156]7H). Furthermore, we
   applied HE staining to examine synovitis severity, Safranin O/Fast
   Green to evaluate the proteoglycan stained by Safranin-O (red) in
   cartilage matrix, and alcian blue staining to check stain acidic
   polysaccharides in cartilages. These staining results indicated a
   significant alleviation of synovitis and cartilage wear in both DMM and
   senile models after injection with FGF18 and LNP-FGF18 mRNA, with lower
   synovitis scores and OARSI scores compared to DMM and senile groups.
   Notably, the LNP-FGF18 mRNA group induced thicker cartilage compared to
   FGF18 protein group (Figs. [157]7F and I-L and [158]8C-G, Supplementary
   Figs. [159]10–[160]11).

   Since we have shown that FGF18 is able to protect cartilage cells from
   degeneration via enhancing autophagy, we evaluated the expression
   levels of Ki67 and LC3 in chondrocytes after the injection of FGF18 and
   LNP-FGF18 mRNA. Local injection of LNP-FGF18 mRNA generate more FGF18
   positive cells, increasing by 2-fold in chondrocytes of knee joint in
   DMM model (Fig. [161]7F, M) and Senile model (Fig. [162]8C, H). FGF18
   mRNA group has significantly more LC3 and Ki67 positive cells in DMM
   model (Fig. [163]7F, N, O) and Senile model (Fig. [164]8C, I, J).

Discussion

   Osteoarthritis (OA) is a degenerative joint disease that is caused by
   multiple factors. Although many biomolecules have shown therapeutic
   potential for the treatment of OA, their efficacy is often limited by
   the depth of penetration and their lifespan in cartilage cavity. In
   this study, we discovered superior penetrating capability of LNP in
   cartilage, which can deliver the mRNA to the deeper zone of the
   cartilage compared to the protein alone. This system is capable of
   achieving chondrocytes in the deeper zones about 50 μm beneath mice
   cartilage, which is superior to recombinant proteins by about 5-fold
   (Fig. [165]2). By optimizing the untranslated region and chemical
   modification of FGF18 mRNA, we constructed a synFGF18 mRNA delivery
   system that can stably express therapeutic doses of FGF18 protein and
   sustain expression in the knee joint cavity of mice for more than 6
   days. LNP-FGF18 mRNA can alleviate the degeneration and senescence of
   chondrocytes induced by inflammatory factors in vitro, while promoting
   chondrocyte proliferation (Figs. [166]4 and [167]5). Mechanistically,
   LNP-FGF18 mRNA can promote FOXO3a expression and activate chondrocyte
   autophagy (Fig. [168]6). We established two OA models, i.e.,
   DMM-induced OA and aging induced OA, to evaluate their efficacy. The
   injection of LNP-FGF18 mRNA into knee joint cavity of mice
   significantly alleviated symptoms of osteoarthritis, such as osteophyte
   formation, cartilage damage, synovial swelling, and decreased pain
   threshold (Figs. [169]7 and [170]8). Due to the advantages of LNP-FGF18
   mRNA over FGF18 recombinant protein, such as greater penetration depth
   and longer lifespan, our in vivo animal experiments demonstrated that
   LNPs had superior effects compared to recombinant proteins, including
   osteophyte number, cartilage thickness, Ki67-positive cell proportion,
   Von Frey test results, and hotplate test results. These results
   indicate the clinical potential of LNP-FGF18 mRNA for the OA treatment.

   Currently, mRNA research focuses on infectious diseases and cancer
   vaccines [[171]12, [172]47], which only need a small amount of antigen
   protein to activate the immune response [[173]18]. However, protein
   replacement therapies, which need recombinant proteins to substitute
   the deficient protein, high level of protein expression is required,
   which is 50-1000 times higher than in vaccine applications [[174]13].
   Further improvements in the mRNA technology are necessary to reduce its
   immunogenicity and improve its translation efficiency [[175]48]. We
   replaced the uridine bases of mRNA with N1-methyl pseudouridine (m1ψ),
   performed Cap1 capping and Poly(A) tailing on the mRNA and optimized
   the UTR segments of the mRNA. These optimizations improved the
   translation efficiency of FGF18 protein by 1000 times compared to the
   natural FGF18 mRNA. In contrast to the relatively short lifetime of
   recombinant proteins [[176]9], in vivo bioluminescence results
   demonstrated that the LNP-mRNA system could locally express the protein
   in the knee joint cavity for up to one week, thus surpassing the
   therapeutic efficacy of one-injection FGF18 protein in mouse models.
   Additionally, the LNP-mRNA system demonstrated good biocompatibility
   and low immunogenicity, with no significant inflammatory response
   observed in the mice’s visceral organs or local synovial tissue.

   Although mRNA requires a carrier to extend its lifetime, the carrier
   offers the potential to overcome certain drug delivery barriers, such
   as penetration depth. The dense, negatively charged extracellular
   matrix (ECM) generates a barrier for traditional drugs to reach the
   chondrocytes. The collagen fibril network (mostly type II collagen with
   some type IX and XI collagen) has an approximate pore size of 60–200 nm
   [[177]10], in consistence with our SEM analyses (Fig. [178]2F), which
   is comparable to the diameter of our LNP systems. Although the collagen
   network in aged mice is denser and has smaller pore sizes compared to
   young mice due to calcification and increased cross-linking of collagen
   molecules that occurs as a result of AGE (advanced glycation end-
   products)-formation [[179]49], it remains the same in diameter to
   prepared LNPs. Our findings suggest that compared to recombinant
   protein, our LNP system can penetrate deeper into chondrocytes,
   resulting in superior therapeutic effects. Further development in
   LNP-mRNA platform, such as lipid structure [[180]50], composition and
   lipid ratios, will improve the penetration and transfection efficiency.

   Compared to other disease, knee joint is a sealed cavity, so it has
   minimal off-tissue side effects. In vivo imaging results have shown
   that only a small amount of LNP leaks out of the joint cavity and
   undergoes hepatic metabolism. After 48 h, there is no detectable signal
   (Supplementary Fig. [181]4). However, in addition to chondrocytes,
   other cell types such as macrophages, synovial cells, and fibroblasts
   exist in the knee joint cavity. LNP-FGF18 mRNA system will also enter
   other types of cells and produce FGF18 protein. However, these proteins
   face the same issue as recombinant proteins, which are difficult to
   reach deep cartilage cells. Additionally, FGF18 mRNA does not only acts
   chondrocytes but also affects other cells. Kuang et al.’s study
   [[182]26] has shown that the lack of FGF18-FGFR3 signaling can enhance
   the CXCL12-dependent chemotaxis of macrophages by upregulating CXCR7,
   thereby exacerbating joint damage in mice. Therefore, the LNP-FGF18
   mRNA system may alleviate arthritis by inhibiting macrophage chemotaxis
   and other mechanisms, which will be investigated in the following work.

   Previous studies have shown that recombinant FGF18 protein can reverse
   degeneration, apoptosis, and reduce intracellular ROS production
   induced by the inflammatory factor IL-1β [[183]36]. However, little
   attention has been paid to the effect of FGF18 protein on chondrocyte
   senescence. This study discovered, for the first time, a significant
   decrease of FGF18 expression in the knee joint cartilage of naturally
   aging (24-month-old) mice. We further demonstrated that FGF18 protein
   and FGF18 mRNA can effectively rescue cellular senescence induced by
   inflammatory factors. Furthermore, we injected the LNP-mRNA into knee
   joints of mice with naturally aging osteoarthritis. Results showed that
   FGF18 mRNA not only alleviated DMM-induced post-traumatic
   osteoarthritis model but also improved local inflammation and
   senescence in naturally aging osteoarthritis, which simulates the
   characteristics of most patients in clinical settings. This discovery
   may expand the use of FGF18 in age-related ailments across multiple
   systems.

   In terms of molecular mechanism of FGF18, traditional studies have
   suggested that FGF18 activates downstream MAPK and Pi3k-Akt pathways by
   binding to FGFR3, thereby promoting chondrocyte proliferation. However,
   this cannot explain the anti-inflammatory and anti-senescence effects
   of FGF18, because MAPK and Pi3k-Akt signaling pathways are already
   overactivated in the inflammatory and senescence environment [[184]51,
   [185]52]. This phenomenon suggests a more comprehensive exploration of
   the FGF18 mechanism of action. Cellular autophagy is a self-protective
   mechanism in which cells form autophagosomes to encapsulate damaged
   organelles, which then fuse with lysosomes for degradation, thereby
   maintaining cellular homeostasis [[186]53]. Autophagy is inhibited in
   osteoarthritis and senescent fibroblast-like synoviocytes [[187]40,
   [188]46]. Our study found that addition of LNP-FGF18 mRNA system
   significantly increased the mRNA level of FOXO3a based on transcriptome
   sequencing results. FOXO3a is a well-known gene and an upstream gene of
   autophagy [[189]54, [190]55]. Further experiments using western
   blotting, immunofluorescence, and transmission electron microscopy
   confirmed that FGF18 can reactivate the autophagy that is suppressed by
   inflammation, restoring normal autophagy levels. The increase
   autophagosomes engulf aging and damaged organelles, thereby alleviating
   cellular inflammation and aging.

   In conclusion, this study developed the LNP-FGF18 mRNA system to
   penetrate the cartilage and stably express FGF18 protein in
   chondrocytes. This system promotes the expression of FOXO3a protein
   that induces autophagy, ultimately reducing inflammation and
   senescence. The LNP-FGF18 mRNA system can slow down the occurrence and
   progression of osteoarthritis both in vitro and in vivo. This strategy
   holds promise as a new therapeutic strategy for OA.

Methods

Human knee cartilage specimen collection

   The sample collection procedure was approved by the ethics committee of
   Shanghai Ninth People’s Hospital affiliated to Shanghai Jiaotong
   University School of Medicine (SH9H-2021-T401-4). Human cartilage
   samples were collected from patients who provided signed informed
   consent. During total knee arthroplasty (TKA) procedures, samples were
   collected from human tibial plateaus. Osteochondral plugs with a size
   of 0.5 cm × 0.5 cm were drilled from the damaged (OA) or intact
   (relatively healthy) parts of the tibial plateau. These removed plugs
   were then fixed in 4% paraformaldehyde for subsequent histological
   assessment using Safranine O/Fast Green staining and FGF18
   immunohistochemistry.

Chemically modified mRNA (mRNA) synthesis and formulation

   mRNA encoding FGF18, Fluc and EGFP gene were synthesized with a T7 RNA
   Polymerase in vitro transcription (IVT) kit bought from Yeasen, China
   (10673ES). The UTR information was listed in the Supplementary Table
   [191]2. During mRNA transcription, uridine was replaced by N1-methyl
   pseudouridine (m^1ψ). IVT reaction details, capping and tailing of mRNA
   and subsequent purification of mRNA was performed according to the
   protocols of kits bought from Yeasen, China (10673ES) and Beyotime,
   China (R7075). Concentration of mRNA were quantified by Nanodrop
   spectrophotometers (Thermo Fisher Scientific, USA). A chimeric UTR is
   designed based on the globin and gp130 associated protein, which is
   named as synUTR.

Lipid nanoparticles production and characterization

   Lipid nanoparticles were composed of four different lipids including
   SM-102, 1,2-distearoyl-sn-glycero-3-PC (1,2-DSPC), cholesterol and
   1,2-dimyristoyl-rac-glycero-3-methoxypolyethylene glycol 2000
   (DMG-PEG2000) bought from Cayman Chemical (LNP-102, USA) at lipid molar
   ratios of 50:10:38.5:1.5 respectively, dissolved in absolute ethanol.
   mRNA was dissolved in 50mM sodium acetate, pH 5.0, and mixed with
   ethanolic lipid mixture with an ethanol: aqueous ratio of 1:3. Mixing
   procedure was performed with a microfluidic device with a Y-channel
   design. After mixture, LNPs were dialyzed in neutral buffer against
   1000 volumes of buffer overnight. Encapsulation efficiency was more
   than 90% for all LNP-mRNAs in this study determined by Ribogreen.
   Particle size and zeta potential of LNPs were evaluated using dynamic
   light scattering (DLS) instrument from Malvern Instruments Ltd. (ZS-90)
   and Bettersize Instrument (BeNano 90). A certain amount of sample was
   dispersed in deionized water (0.1% w/v), followed by sonication for
   3 min. The measurements were performed three times, and the average
   values were calculated.

Cryo-transmission electron microscopy (Cryo-TEM), TEM and scanning electron
microscopy (SEM) analyses

   For cryo-TEM, a polymer-coated carbon-reinforced copper grid was
   treated with a small drop (approximately 3 μl) of mRNA-LNPs solution
   for 1 min, blotted for 5 s, and swiftly plunged into liquid ethane
   cooled by liquid nitrogen. This process was carried out at 8 °C and
   100% humidity using a Vitrobot Mark IV from FEI, Thermo Fisher
   Scientific. The grids were subsequently imaged with a Talos Arctica
   system at 200 kV, utilizing a Falcon 3EC detector and EPU for data
   collection.

   For TEM, the samples were fixed with 2.5% glutaraldehyde and kept at
   4 °C for 12–24 h. Subsequently, fixed samples were dissolved in 1%
   osmium tetroxide for 1–2 h and washed three times with PBS for 15 min
   each time. The samples were then dehydrated and infiltrated with a
   gradient concentration of acetone for embedding. The resin blocks were
   cut into 70–90 nm ultrathin sections using an ultramicrotome, followed
   by retrieval of the sections using copper grids. Sections were stained
   with uranyl acetate for 8–10 min and lead citrate for 8–10 min. After
   drying, the samples were observed and images were captured for analysis
   using a transmission electron microscope. For SEM, samples were fixed,
   dehydrated and coated with 10 nm of gold before SEM scans (Gemini300,
   Carl Zeiss).

Cartilage infiltration assay

   Recombinant FGF18 was labeled by Cy3 (BRC-15, Conflure Biology) with
   protocol “NHS Ester Labeling of Amino-Biomolecules,
   [192]www.lumiprobe.com”. In short, Cy3 NHS ester was mixed with FGF18
   in a 0.1 M sodium bicarbonate buffer and incubated in dark for 8 h. LNP
   was labeled with Dil (MedChemExpress, HY-D0083) solution with a
   concentration of 100 μM and incubated for 8 h in the dark. Cy3-FGF18
   and Dil-LNPs were purified using size exclusion chromatography columns
   to remove free dyes (Thermo Fisher Scientific, 89882). For cartilage
   infiltration assay, Cy3-FGF18 and Dil-LNPs were intra-articularly
   injected to young (3 m) and old (18 m) mice. Knee joints were collected
   after 24 h for subsequent frozen sections. Nuclei was stained with DAPI
   and images were taken with a confocal microscopy.

Animals

   All animal experiments were conducted in accordance with the guidelines
   approved by the ethics committee of the Shanghai Ninth People’s
   Hospital affiliated to Shanghai Jiaotong University School of Medicine
   (SH9H-2023-A856-1). C57BL/6J mice of wild-type were obtained from the
   animal center of Shanghai Ninth People’s Hospital. The mice were housed
   in specific pathogen-free (SPF) cages under standard conditions, with a
   temperature of 24℃, a humidity of 60%, and a 12-hour light-dark cycle.
   They were provided with a standard diet.

   For the DMM surgery, 10-week-old C57BL/6J mice were selected and
   anesthetized with isoflurane. The right knee joint was exposed, and the
   medial meniscotibial ligament was transected to induce destabilization
   of the medial meniscus, following a previously established protocol
   [[193]56]. 500ng FGF18 protein (20 μL, 25 μg/mL) or 6 μg LNP-FGF18 mRNA
   (20 μL, 300 μg/mL) was injected intraarticularly in the right knee once
   a week after surgery. After eight weeks post-surgery, the mice were
   euthanized, and the right knee joint was collected. The joints were
   fixed in 4% paraformaldehyde (PFA) for 48 h and then transferred to 75%
   ethanol for subsequent micro-computed tomography (micro-CT) and
   histological analysis.

   For the senile osteoarthritis (OA) model, C57BL/6J mice were fed a
   normal diet and allowed to age until 24 months, with 3-month-old mice
   serving as controls. 500 ng FGF18 protein (20 μL, 25 μg/mL) or 6 μg
   LNP-FGF18 mRNA (20 μL, 300 μg/mL) was injected intra-articularly in
   knee joints of 18-month-old mouse once a week for two months. Right
   lower extremities were collected and fixed for further radiological and
   histological analysis.

Gait analyses

   Gait analyses was evaluated with the CatWalk gait analysis system
   (Noldus Technology, Netherlands) after eight weeks post-surgery or
   senile mice at 24-month-old [[194]57]. Mice were pretrained to get used
   to crossing the walkway for thirty minutes before gait recording. The
   recordings were carried out in a dark environment. Gait analysis was
   recorded using a video-based system. The paw, upon contacting the glass
   plate, reflected green light to a video camera, which recorded the
   entire run at a frequency of 100 Hz. The same investigator conducted
   the entire gait analysis experiments with blindness to grouping. Both
   hind and front paws were assessed for data analysis. Compliant runs
   were defined with a maximum speed variation of 60%. Runs with backing,
   sniffing, cleaning, or long stops were excluded. Three compliant runs
   were required for statistical analysis. CatWalk software will
   automatically label all areas and identify and assign them to the
   respective paw. The following parameters were analyzed: max contact
   area, print area, mean intensity, and duty cycle (percentage of stance
   phase in a step cycle).

Von-frey test and hotplate test

   The Von Frey test was conducted to measure the pain threshold in the
   hindlimbs of mice using an electronic Von Frey filament (Xinruan
   Technology, China). Prior to the test, each mouse was allowed to
   acclimate to the test chamber for 30 min. A positive response was
   recorded if the animal exhibited any nociceptive behavior, such as paw
   withdrawal, shaking, or licking. The threshold force exerted by the
   electronic filament was recorded, with at least three replicative
   forces measured for each mouse in each test. The Von Frey test was
   performed every two weeks after the DMM surgery.

   For hotplate test, mice were subjected to the hotplate test at a
   temperature of 55 °C. The response time was recorded as the duration
   between the hindlimbs touching the plate and response behaviors such as
   paw shaking, paw licking, or jumping. The hotplate pain assay was
   performed every two weeks following the DMM surgery. At least three
   replicative response times were recorded for each mouse. Both Von Frey
   or hotplate tests were conducted by an observer who was blinded to the
   grouping of mice.

Micro-CT

   Knee joints from mice were fixed for two days and then scanned using a
   high-resolution micro-CT scanner (Skyscan 1072, Belgium) with a pixel
   size of 9 μm, a 55 kVp source, and a 145 μAmp current. Osteophyte
   volume, bone volume/tissue volume (BV/TV), trabecular number (Tb. N),
   trabecular thickness (Tb. Th) and trabecular separation (Tb. Sp) of
   subchondral bone were calculated with CTAn software (Bruker, Germany).
   Three-dimensional reconstructions of subchondral bone sections were
   produced with CT-Vox software (Bruker, Germany), as described in our
   previous study [[195]58].

Histological analysis

   For histologic analysis, knee joints and human specimens were fixed,
   decalcified, dehydrated, and embedded into paraffin. Serial tissue
   sections of a thickness of 5 μm were cut in a sagittal or coronal
   plane. Safranine O/Fast Green staining was performed to assess
   cartilage degeneration of tibial and femur cartilage using the
   Osteoarthritis Research Society International (OARSI) scoring system
   [[196]59]. Hematoxylin and eosin (HE) staining was also performed to
   evaluate synovial inflammation using the synovitis score [[197]60,
   [198]61].

   For immunohistochemistry (IHC) staining, the sections were incubated
   with a primary antibody against FGF18 (11495-1-AP, 1:200, Proteintech),
   Ki67 (1:200, ab16667, Abcam) and LC3 (14600-1-AP, 1:100, Proteintech).
   Images were captured using a Zeiss DM4000B high-resolution microscope,
   and quantitative analysis was performed using the ImageJ software.
   Positive cells were counted by two different researchers (KK and BL).
   Average number of positive cells calculated by the two researchers was
   used as the final count of positive cells. This number was then divided
   by the total number of cells in the section to determine the proportion
   of positive cells.

In vivo bioluminescence

   On days 1, 2, 3, 4, and 6 after injecting LNP-Fluc mRNA into the right
   knee joint cavity of C57BL/6J mice, bioluminescent imaging was
   performed with an IVIS Spectrum system (PerkinElmer, USA). To prevent
   interference from dark-colored fur, the hair on the lower limbs of the
   mice was removed. D-luciferin was diluted in phosphate-buffered saline
   (PBS) and injected intraperitoneally at a dose of 150 mg/kg. After
   20 min, the mice were anesthetized using isoflurane. Photon emissions
   from the right lower limb were collected with a 1-minute exposure time.
   Luminescent signal was overlaid onto the mouse photo, and the regions
   of interest were delineated using Living Image software (PerkinElmer,
   USA) to calculate the signal flux within the regions.

   For assessing the metabolic pathway of LNP-mRNA, bioluminescent imaging
   was performed on mice 24 and 48 h after injecting LNP-Fluc mRNA into
   the knee joint cavity of C57BL/6J mice. D-luciferin was injected
   intraperitoneally at a dose of 150 mg/kg. After 20 min, the mice were
   euthanized, and their hearts, livers, spleens, lungs, and kidneys were
   collected and imaged using the IVIS imaging system with a 1-minute
   exposure time. Living Image software was used to detect the signal flux
   within the regions of interest.

Cell culture and reagents

   Murine primary chondrocytes were isolated from newborn C57BL/6J mice
   following a previously established protocol [[199]62]. Primary
   chondrocytes were cultured in Dulbecco’s Modified Eagle Medium/Nutrient
   Mixture F12 (DMEM/F12) with 10% fetal bovine serum (FBS) and 1%
   penicillin-streptomycin. The ATDC5 cell line, an immortalized mouse
   chondrocyte cell line, and HEK293T cells were obtained from the Cell
   Bank of the Chinese Academy of Sciences (Shanghai, China). ATDC5
   chondrocytes were cultured in DMEM with 4.5 g glucose/L, 5% FBS, and 1%
   penicillin-streptomycin. 293T cells were cultured in DMEM with 4.5 g
   glucose/L, 10% FBS, and 1% penicillin-streptomycin. IL-1β and TNF-α
   protein was purchased from Genscript. Recombinant human FGF18 protein
   was purchased from Sino Biological (13206-H08H). 3-Methyladenine (3-MA)
   was purchased from TargetMol (T1879). Lipofectamine 3000 was purchased
   from Thermo Fisher Scientific, United States. 10 ng/mL IL-1β, 10 ng/mL
   TNF-α and 5 mM 3-MA were used to incubate chondrocytes.

Dot blot

   Dot blot assay was used to evaluate the protein expression of FGF18 in
   cell culture supernatants. After transfecting ATDC5 or 293T cells with
   LNP or plasmid for 24 h, cell culture supernatants were collected and
   centrifuged at 3000 rpm for 15 min to obtain the supernatant. Then,
   30 μl of the supernatant was spotted onto a nitrocellulose membrane,
   and 10 μg and 1 μg of rhFGF18 were spotted as positive controls. After
   complete absorption of the liquid on the membrane, it was blocked with
   5% BSA (Bovine Serum Albumin) for 60 min. After washing three times
   with TBST (tris-buffered saline and Tween 20), the membrane was
   incubated overnight at 4 °C with the primary antibody against FGF18
   (1:1000, 11495-1-AP, Proteintech). On the following day, after washing
   three times with TBST, the membrane was incubated with the
   corresponding fluorescent secondary antibody at room temperature for
   2 h. The membrane was then visualized using the Odyssey V3.0 image
   scanner (Li-COR. Inc., Lincoln, NE, USA). The fluorescence intensity of
   the supernatant dots and the rhFGF18 dots were measured using
   ImageStudio Lite software (Li-COR) to determine the expression level of
   FGF18 protein in the supernatant.

Luciferase reporter gene assay

   Cells were seeded in a black-bottom 96-well plate (BS-MP-96B, Biosharp)
   and transfected with different concentrations of LNP-Fluc mRNA. At
   different time points after transfection (24 h, 48 h, 72 h), the
   firefly luciferase activity was measured using the firefly luciferase
   reporter gene assay kit (RG051, Beyotime). 100 μl of the assay reagent
   was added to each well of the 96-well plate and incubated at room
   temperature for 5 min. Subsequently, chemiluminescent signals were
   detected using a multifunctional microplate reader.

Plasmid and lentivirus transfection

   FGF18 plasmids with different UTRs were synthesized by Genecefe
   Biotechnology Inc. China. Sh-FOXO3a lentivirus was purchased from Obio
   technology, China. Plasmid was transfected with the protocol of
   Lipofectamine 3000 and the protein expression was detected at 48 h
   after transfection. For lentiviral infection, we added lentivirus at a
   multiplicity of infection (MOI) of 10 and incubated the cells. After
   48 h of infection, the media was replaced, and the successfully
   transfected cells were selected using puromycin to establish a stable
   FOXO3a knockdown cell line.

Micromass culture

   After digestion, centrifugation, and suspension, a suspension of ATDC5
   chondrocytes was prepared at a concentration of 1.5 × 10^^7 cells per
   mL. Then, 10 μL of the cell suspension was added to the center of each
   well in a 24-well culture plate. The plate was incubated in a cell
   incubator for two hours to allow cell attachment. 1 mL of cell medium
   containing 1% insulin-transferrin-selenium (ITS) (41400045, Thermo
   Fisher) and either 100 ng of FGF18 protein or 3 μg of LNP-FGF18 mRNA
   was gently added to each well. Cell medium was changed every two days.
   After different time periods, the cell masses were fixed and stained
   with Alcian blue or toluidine blue dyes for 30 min. The integrated
   optical density (IOD) value of each micromass was calculated using
   Image J software.

Cell toxicity

   To test the toxicity of the LNP-mRNA system on cells, we used cell
   counting kit-8 (cck8) (Dojindo, Japan) and calcein/pi staining kit
   (C2015, Beyotime, China) to test the toxicity of LNP. In the cck8
   experiment, we treated cells with different concentrations of LNP
   system for 24 h, 48 h, and 72 h. Then, at different time points, we
   added 10% cck8 reagent (Dojindo, Kumamoto, Japan) to the cells and
   incubated them in a cell incubator for 2 h. Finally, an Infinite M200
   pro multimode microplate reader (Tecan Life Sciences, Switzerland) was
   used to detect the absorbance at 450 nm of each well. For calcein/pi
   staining, we added calcein/pi working reagent to cells and incubated
   them at 37 °C in the dark for 30 min. Then calcein and pi positive
   cells were observed and quantified under a laser confocal microscope.

EdU assay

   EdU assay was utilized to assess cell proliferation by incorporating
   the thymidine analogue, EdU (5-ethynyl-2’-deoxyuridine), into DNA
   replication. The EdU assay protocol followed the instructions provided
   with the kits purchased from Beyotime, China (C0075). Primary
   chondrocytes were incubated with a 10 μM EdU working solution at 37 °C
   for 2 h to label the cells with EdU. Following this, the cells were
   fixed, washed, and permeabilized. To prepare the click reaction
   solution, we combined components such as click reaction buffer, CuSO4,
   azide 555, and click additive solution. The prepared Click reaction
   solution was then added to the cells and incubated in the dark for
   30 min. Finally, we captured representative images using a laser
   confocal microscope and further quantitative analysis was done with
   Image J.

Senescence-associated-β-galactosidase (SA-β-gal) staining

   Cellular senescence was tested with a SA-β-gal staining kit bought from
   Beyotime, China (C0602). Primary chondrocytes were cultured with FGF18,
   LNP-FGF18 or IL-1β for 7 days. The SA-β-gal staining procedure was
   performed following the instructions provided with the kit. Random
   microscopy fields were selected for image acquisition. ImageJ software
   was utilized to count the number of β-gal staining-positive cells for
   subsequent quantitative analysis.

Immunofluorescence (IF)

   For cell immunofluorescence, we treated ATDC5 chondrocytes with
   different drugs for 24 h. Then, we fixed, permeabilized, blocked, and
   washed the cells. Next, we incubated the cells with the corresponding
   primary antibodies overnight at 4 °C. After that, we incubated the
   cells with Alexa Fluor-conjugated secondary antibodies (1:1000) at room
   temperature for 1 h and stained the cell nuclei with DAPI. Finally, we
   observed and captured representative images under a laser confocal
   microscope and quantified the number of positive cells using ImageJ
   software. COL2A1 antibody (1:100, ab34712), SOX9 antibody (1:100,
   ab185230), and Ki67 antibody (1:200, ab16667) were all purchased from
   Abcam.

Western blot

   After treating cells with different conditions for 24 h, we lysed the
   cells using RIPA lysis buffer (P0013, Beyotime) supplemented with 1%
   protease/phosphatase inhibitor and purified and quantified cell
   proteins through centrifugation and bicinchoninic acid assay (BCA).
   20 μg protein was loaded into each well of SDS-PAGE gel, and
   Tris-Glycine was used as the electrophoresis buffer. After
   electrophoresis at 140 V for 1 h, protein was then transferred to a
   PVDF membrane and blocked with a 5% BSA solution. Subsequently, the
   membrane was incubated with the corresponding primary and secondary
   antibodies and the protein bands were imaged using an Odyssey V3.0
   image scanner (Li-COR. Inc., Lincoln, NE, USA). COL2A1 antibody
   (1:1000, ab34712), SOX9 antibody (1:1000, ab185230), MMP3 antibody
   (1:1000, ab52915), and MMP13 antibody (1:1000, ab39012) were purchased
   from Abcam. β-actin antibody (4970, 1:2000), BECLIN1 antibody (3495,
   1:1000), FOXO3a antibody (12829, 1:1000), and LC3B antibody (43566,
   1:1000) were purchased from CST. FGF18 antibody (11495-1-AP, 1:1000)
   were purchased from Proteintech.

RNA sequencing

   Chondrocytes were seeded in six-well plates at a density of 4 × 10^6
   cells per well. After 24 h of incubation with 10 ng/mL TNF-α (C600052,
   Sangon) or 3 μg/mL LNP-FGF18 mRNA, total RNA was extracted using the
   Total RNA Extraction Kit (R6812-01HP, Omega, United States). RNA
   sequencing was performed by Wuhan Huada Gene Technology Co., Ltd.
   (China). Gene Ontology (GO) enrichment analysis, Kyoto Encyclopedia of
   Genes and Genomes (KEGG) pathway enrichment analysis, and gene set
   enrichment analysis (GSEA) were conducted using the Mybgi platform
   (Wuhan Huada Gene Technology, [200]https://mybgi.bgi.com/tech/login).
   Heatmap analysis was performed in GraphPad Prism 8.0 (GraphPad Software
   Inc., San Diego, CA, USA) based on the Z-score of TPM counts for each
   gene.

Quantitative real-time PCR (qRT-PCR)

   Total RNA was extracted from cells using a Total RNA Extraction Kit
   (R6812-01HP, Omega, United States). 1000 ng cDNA was synthesized with
   the aid of a cDNA synthesis kit (RR036A, Takara, Japan), diluted in 200
   μL of ddH[2]O. A total of 10 μL of the qRT-PCR mix, consisting of 3.6
   μL of double distilled H[2]O, 1 μL of cDNA, 0.2 μL of upstream and
   downstream primers respectively, and 5 μL of SYBR premix ([201]B21202,
   Selleck, USA), was added to 384-well plate. qRT-PCR process was
   performed on an ABI 7500 Sequencing Detection System (Applied
   Biosystems, USA). The primer sequences used for the target genes are
   listed in Supplementary Table [202]1.

Statistical analysis

   Statistical analyses were conducted using SPSS version 25.0 (SPSS Inc.,
   Chicago, IL, USA). For data that followed a normal distribution,
   comparisons between two groups were performed using the unpaired
   Student’s t-test. One-way ANOVA followed by Turkey’s post hoc tests
   were used for comparisons among three or more groups. Non-normally
   distributed data were analyzed using the Mann-Whitney and
   Kruskal-Wallis tests. GraphPad Prism 8.0 (GraphPad Software Inc., San
   Diego, CA, USA) was utilized for the design and creation of statistical
   charts. All P values reported in this study are two-sided, and
   statistical significance was defined as P < 0.05, **P < 0.01,
   ***P < 0.001, and ****P < 0.0001.

Electronic supplementary material

   Below is the link to the electronic supplementary material.
   [203]Supplementary Material 1^ (5.5MB, docx)

Acknowledgements