Peptide-based targeted drug delivery strategies for osteoarthritis treatment

Abstract

Peptide-based strategies offer promising solutions to overcome the complex, multi-tissue barriers of osteoarthritis. Their tunability, specificity, and versatility enable targeted drug delivery to cartilage, synovium, and subchondral bone, while some therapeutic peptides provide intrinsic anti-inflammatory, regenerative, or analgesic effects. Advances in peptide design, stability engineering, and in silico screening, alongside emerging human joint-on-chip models, are accelerating the development of targeted, stable, multi-tissue OA therapies.

Introduction

Osteoarthritis (OA) remains without a cure, owing to the dual challenges of its complex, multi-tissue pathogenesis and the challenging barriers to effective drug delivery within the joint. OA is now recognized as a whole-joint disease, driven by dynamic crosstalk among multiple tissues, each exhibiting distinct pathological phenotypes¹. While cartilage degradation and synovial inflammation are commonly addressed hallmarks of OA, other joint components, including subchondral bone (SB), fat pad, ligament, meniscus, and synovial fluid (SF), also contribute critically to disease progression and must be addressed for comprehensive intervention². These diverse tissues create unique physicochemical and structural barriers that complicate the delivery of therapeutics, whether by intra-articular (IA) or systemic routes (Fig. 1). IA administration offers improved local bioavailability and reduced systemic exposure compared to systemic dosing. However, its pharmacokinetics are limited by rapid joint clearance, competitive binding with SF, enzymatic degradation, and restricted penetration through dense extracellular matrices (ECMs)³. Conversely, systemic administration enables delivery to vascularized tissues, including the synovium and subchondral bone, but suffers from poor joint specificity, enzymatic degradation, and rapid systemic clearance³. Together, these challenges underscore the pressing need for drug delivery systems capable of achieving efficient, tissue-specific distribution and retention within the joint.

Fig. 1. Challenges in drug delivery for OA treatment.

Therapeutics for OA can be administered via IA or systemic injection, depending on the target tissue. IA delivery enables localized treatment of avascular cartilage but faces challenges including enzymatic degradation, competitive binding with SF constituents, rapid clearance through capillary and lymphatic drainage, and hindrance by the dense extracellular matrix, limiting tissue penetration (red arrow). Systemic administration can target vascularized tissues such as synovium and subchondral bone yet minimizing off-target delivery to other organs, maintaining drug stability in circulation, and achieving targeted joint accumulation remain major obstacles (green arrow). These challenges severely limit the bioavailable drugs delivered to the target tissue for therapeutic benefits (black arrow).

Among emerging strategies, peptides offer a uniquely versatile platform for targeted drug delivery. Their tunable amino acid sequences can be rationally designed to exhibit high affinity and specificity toward distinct tissue or cellular targets^4,5. Positioned between small molecules and large biologics, peptides strike an ideal balance: offering superior surface interactions and stability relative to small molecules, while avoiding immunogenicity, limited tissue penetration, and high production costs that often constrain biologics. Their modularity further enables integration into diverse formulations like targeting ligands or carriers, and in some cases, they can function as therapeutics themselves. Collectively, these attributes highlight peptides as a powerful and adaptable class of agents for overcoming joint transport barriers and advancing next-generation OA therapies⁶. This review focuses on peptides for OA treatment, specifically those that enable targeted drug delivery to distinct joint tissues and possess intrinsic therapeutic activity addressing key OA pathophysiology mechanisms, including impaired anabolism, inflammation, and pain.

Discovery and synthesis

Two complementary strategies underpin the discovery of targeting peptides: library-based screening and rational design (Fig. 2). Library-based screening involves screening/biopanning of phage display’s random peptide library or well-defined peptide library against the target of interest. The majority of peptides used to target cartilage, synovium, and SB were identified through this technique. In contrast, rational design relies on mechanistic understanding of binding interactions to engineer peptides with defined physicochemical or structural features. This strategy often exploits dominant biophysical properties of the target tissue—such as charge, hydrophobicity, or matrix affinity—or adapts motifs derived from binding domains of natural proteins.

Fig. 2. Discovery of targeting and therapeutic peptides.

Targeting peptides can be identified through library-based screening or rational design. In library-based screening, random or predefined peptide libraries are screened against ECM components, cells, tissues, or in vivo systems to isolate peptides with high affinity and specificity. Phage display is the most widely used method, in which bacteriophages displaying random peptides undergo iterative biopanning against a target. Rational design leverages dominant biophysical features—such as the anionic GAG network of cartilage or the cationic HAP of bone—or adapts known binding motifs from natural proteins (e.g., the heparin-binding domain of porcine EGFR or repetitive residues from bone binding proteins) to confer tissue-specific affinity for the peptide. Therapeutic peptides can be identified by analyzing differential protein expression between healthy and osteoarthritic cartilage, followed by interaction-network analysis to pinpoint candidate sequences. Additionally, drugs used in non-joint indications may be repurposed for OA interventions.

Therapeutic peptides for OA are frequently inspired by endogenous bioactive peptides that possess intrinsic anti-inflammatory, regenerative, or analgesic effects. Candidate sequences are often identified through comparative analyses of protein expression and molecular interaction networks involved in OA pathogenesis, which help pinpoint peptides with potential disease-modifying activity (Fig. 2). Following their discovery, optimization typically focuses on enhancing peptide pharmacokinetics and stability further elaborated in the next section.

Two common methods are used for peptide synthesis: solid-phase peptide synthesis (SPPS) and recombinant expression. SPPS remains the dominant method, providing precise control over sequence composition through iterative coupling and deprotection of amino acids on a solid resin support, followed by cleavage of the completed peptide from the resin. This approach is highly efficient for synthesizing short to medium-length peptides (typically < 50 residues) and allows the incorporation of noncanonical amino acids or chemical modifications. In contrast, recombinant expression leverages the cell’s translational machinery and is generally preferred for producing long peptides or peptide-protein fusions, where SPPS becomes inefficient. Both approaches permit the incorporation of modifications, some of which are method-specific, to improve peptide targeting and therapeutic performance.

Application: targeting peptide

The joint provides multiple tissue targets for OA therapy, including cartilage, synovium, and SB, which collectively drive disease progression. Targeting peptides can be incorporated into drug delivery systems through diverse strategies: covalent conjugation to therapeutic payloads, complex formation as peptide-payload assemblies, genetic fusion with protein therapeutics, or surface/scaffold functionalization of nano or microscale carriers.

Cartilage

Cartilage is essential for joint biomechanics, providing a low-friction, load-bearing surface. In OA, elevated pro-inflammatory and catabolic mediators, coupled with reduced anabolic activity, lead to structural deterioration, owing to ECM degradations, and chondrocyte dysfunction (Fig. 3a). Chondrocytes, being the sole resident cells, represent prime therapeutic targets yet are difficult to access due to their sparse distribution within a dense ECM. Conversely, the abundance of ECM constituents, such as collagens, aggrecan, and hyaluronic acid (HA), presents multiple accessible binding targets (Table 1).

Fig. 3. Cartilage: tissue phenotype in OA and examples of targeting peptides.

a Schematic illustration of healthy versus OA cartilage, highlighting key structural and compositional alterations and potential molecular targets within the tissue. b(i) CPC exhibited strong intra-cartilage accumulation and retention for up to 7 days in a rabbit OA model after IA injection. Reprinted with permission from Elsevier, copyright 2025²⁰. b(ii) CPC-modified exosomes were able to fully penetrate cartilage thickness (left; red: AlexFluor647) and enabled GFP mRNA transfection in a rat OA model following IA injection (right; green: GFP). Reprinted with permission from Wiley, copyright 2024²⁴. c An 8-arm PEG system incorporating COLBP and HABP demonstrates localization to degenerated cartilage of a mouse OA model following IA injection (left: IHC of biotin-tagged peptide, brownish red; right: Safranin O staining). Reprinted with permission from Elsevier, copyright 2018³⁶. d(i) CAP-PEI complexes delivered GFP plasmids to cartilage following IA injection in healthy rat (blue: Hoechst 33258; green: GFP). Reprinted with permission from Elsevier, copyright 2011³⁸. d(ii) CAP-modified exosomes exhibited prolonged joint retention up to 6 days post-IA injection in rat joints, indicated by near-infrared signal. Reprinted with permission from Elsevier, copyright 2024⁴⁰.

Table 1.

Cartilage targeting peptides and their applications

Target	Peptide	Discovery	Application	Outcomes	Citation
Collagen II	WYRGRL; COLBP	In vitro phage display	Carrier functionalization: Functionalized poly(propylene sulfide) nanoparticle	Higher accumulation of nanoparticle in cartilage ECM compared to random peptide modification	9
			Peptide-drug conjugate: Modified tetrapodal DOTAM tethered with cathepsin D inhibitor	Prolonged retention in rat knee for at least 7 days	112
			Peptide-peptide conjugate: HA binding peptide (HABP)-COLBP with multi-arm PEG linker	Reduced the expression of inflammatory genes, pain, and cartilage degradation in OA mice	36
			Peptide-drug conjugate: COLBP-Dexamethasone	Improved dexamethasone efficacy in OA cartilage explant by prolonged intra-cartilage retention	113
			Carrier functionalization: Rapamycin and bilirubin-loaded functionalized nanoparticle	Suppressed inflammation and showed cartilage protection effect in OA rat	10
Collagen XII	DLQYWYPIWDTH; C5-24	In vitro phage display	Peptide-contrast agent conjugate: Superparamagnetic iron oxide (SPIO)-C5-24	Enabled arthritic cartilage detection in OA pig	11
			Peptide-ECM conjugate: HA-C5-24	Optimum cartilage lubrication effect on human OA cartilage explants and OA rat
				Peptide-coated MSCs enhanced tissue repair in an arthritis rat model
Aggrecan	RRRR(NNRRR)3R; CPC + 14 N	Rational Design: electrostatic interactions with GAG	Peptide-drug conjugate: CPC-Insulin-Like Growth factor-1 (IGF-1)	Prolonged protection against matrix degradation in cartilage explant OA model by a single dose	21
			Peptide-drug conjugate: CPC-Interleukin-1 Receptor Antagonist (IL-1Ra)	Prolonged protection against matrix degradation in cartilage explant OA model by a single dose	22
			Carrier functionalization: Surface-modified exosome for IL-1Ra delivery	Enabled IL-Ra delivery to suppress cartilage matrix degradation in cartilage explant OA model	25
			Carrier functionalization: Surface-modified exosome for IL-1Ra delivery	Enabled mRNA delivery to articular chondrocytes of OA mouse	24
	(RRAAAA)3RR; CPC + 8		Peptide-contrast agent conjugate: CPC- ioxaglate (IOX)	Enhanced IOX intra-cartilage transport at lower dosage for CT imaging	28
	GSGVPINVRCRGSR-DCLDPCRRAGMRFGR-CINSRCHCTP; Cysteine Dense Peptide (CDP)	In vivo library screening	Peptide-drug conjugate: CDP-triamcinolone acetonide (TAA)	Localized distribution in cartilage and intervertebral disc (IVD) and reduced inflammation in OA rat	30
	ε-Poly-l-lysine	Rational Design: electrostatic interactions with GAG	Peptide-drug conjugate: Kartogenin-ε-Poly-l-lysine	Improved OA histological score and collagen II expression in OA rat	11
			Carrier functionalization: Surface-modified mesenchymal stem cell-derived small extracellular vesicles	Enhanced intra-cartilage transport, improved cartilage histology, and reduced pain response in OA rat	114
	Poly-lysine; Branched Poly-lysine (BPL)	Rational Design: electrostatic interactions with GAG	Peptide-drug conjugate: BPL-methylprednisolone	Prolonged retention in healthy mouse joint	115
			Peptide-ECM conjugate: HA-BPL	Adhered to cartilage and showed effective cartilage lubrication properties	116
Heparin Sulfate-GAG	KKKRKGKGLG-KKRDPCLKKY	Rational Design: known heparin binding (HB) domain of EGF (epidermal growth factor-like growth factor)	Peptide-protein fusion: HB-IGF-1	Enhanced intra-cartilage retention and cartilage protection effect of IGF-1 in OA rat	117
HA	GAHWQFNALTVR; HABP1	In vitro phage display	Carrier functionalization: Functionalized PEG hydrogel	Enhanced cartilage formation in OA rat	37
			Peptide-peptide conjugate: HABPs-COLBP with multi-arm PEG linker	Reduced the expression of inflammatory genes, pain, and cartilage degradation in OA mouse	36
	RYPISRPRKRC; HABP2	Rational Design: homolog to proteoglycan link protein
	STMMSRSHKTRSHHV	Rational Design: homolog to Receptor for Hyaluronan-mediated Motility (RHAMM) protein	Direct application: peptide-high molecular weight HA (HMWHA) supplementation	Supplementation to IL-1-stimulated chondrocyte culture showed anti-inflammatory effect	118
Chondrocyte	DWRVIIPPRPSA; Chondrocyte Affinity Peptide (CAP)	In vitro phage display	Peptide-payload complex: Polyethylenimine-CAP/plasmid (GFP and luciferase) complex	Enhanced intra-cartilage transport enabling plasmid transfection into chondrocytes of healthy rat	38
			Carrier functionalization: Surface-modified exosome for mir-140, MMP-13 siRNA, and antisense oligonucleotides (ASO) delivery	Targeted cartilage delivery and prolonged joint retention, which enhanced payload efficacy in OA rat	39,40,119

Collagen constitutes roughly two-thirds of cartilage dry weight, with 90–95% being type II collagen, characterized by triple-helix domains rich in polar and hydrophobic residues⁷. Its unique structural motifs enable specific binding, particularly by peptides⁸. The collagen-binding peptide (COLBP; WYRGRL), identified by phage display, integrates charge and hydrophobicity for selective binding⁹. COLBP-decorated poly(propylene sulfide) nanoparticles localized in healthy rat cartilage following IA injection, while COLBP-modified mesoporous polydopamine nanoparticles enabled targeted delivery of reactive oxygen species (ROS) scavengers in rat OA joints¹⁰. Disease-induced ECM remodeling also creates new binding targets (Fig. 3a). The phage display-selected peptide C5-24 (DLQYWYPIWDTH) binds upregulated collagen XII in arthritic cartilage¹¹. When conjugated to superparamagnetic iron oxide, C5-24 enabled MRI visualization of degenerated cartilage in a porcine OA model, and when linked to HA, it enhanced cartilage homing and lubrication versus HA alone.

Aggrecan, the proteoglycan “bottlebrush” densely decorated with glycosaminoglycans (GAGs), is the primary source of cartilage’s high negative fixed charge density¹². This electrostatic environment not only imparts exceptional compressive load-bearing capacity but also drives strong Donnan-mediated uptake of cationic drug carriers¹³. Although aggrecan is among the first cartilage ECM components degraded during OA pathogenesis¹⁴, the remaining aggrecan-GAGs can still exhibit strong Donnan-mediated uptake of cationic carriers, enabling higher intra-cartilage accumulation compared to neutral counterparts^15–17. Cationic peptide carriers (CPCs), arginine-rich peptides engineered for highly distributed positive charge and minimal hydrophobicity¹⁷, exemplify this principle, exhibiting rapid intra-cartilage penetration and prolonged retention^16,18,19. In a rabbit OA model, CPCs achieved full-depth cartilage penetration and remained detectable for up to seven days after a single intra-articular injection²⁰ (Fig. 3b (i)). These enhanced targeting properties of CPCs have been utilized to facilitate the delivery of protein therapeutics for sustained suppression of matrix degradation in a cartilage explant OA model^21,22. CPCs can also be used to functionalize exosomes, imparting a positive surface charge that enhances transport through negatively charged tissues^5,23. CPC-decorated exosomes demonstrated efficient tissue targeting and enabled deep-cartilage GFP mRNA transfection in rat OA joints (Fig. 3b (ii))²⁴, and have been successfully used to deliver therapeutic cargo to cartilage²⁵. Beyond cartilage, CPC-modified exosomes have also enabled delivery to other charge-dense tissues, including mucus²⁶ and vitreous humor²⁷. Additionally, owing to their rapid intra-cartilage penetration, CPCs are also well-suited to enhance imaging applications as exemplified by CPC–IOX which produced equivalent cartilage CT signals at 40-fold lower doses than free IOX²⁸. Beyond CPCs, other cationic systems also exploit charge-based targeting. Cationic ε-poly-L-lysine improved kartogenin efficacy in a rat OA model after IA injection and potentially imparted protease resistance through its noncanonical residues²⁹. In contrast to the rational design of CPC and ε-poly-L-lysine, cysteine-dense peptides (CDPs) were identified from a toxin library screening for peptide with the highest accumulation in cartilage after IV injection³⁰. Their disulfide-stabilized structures preserved positive surface charge and resisted proteolysis in circulation. Intravenous doses of CDP–triamcinolone conjugate reduced arthritic rat joint inflammation while minimizing systemic steroid exposure. Extending beyond peptides, cationic proteins such as GFP and avidin also serve as viable platforms for cartilage-targeted drug delivery^15,31–33.

Hyaluronic acid (HA), another negatively charged ECM component, can be targeted using HA-binding peptides (HABPs), which are generally cationic and enriched in polar residues enabling electrostatic and hydrogen-bonding interactions. HABPs such as GAHWQFNALTVR and RYPISRPRKRC were discovered through phage display³⁴ or derived from link protein sequences³⁵. Their use in OA therapy has shown promise when combined with other targeting motifs or scaffold functional groups. For example, a hybrid HABP-COLBP peptide-polymer localized in degenerated cartilage and other joint tissues of a mouse OA model following IA injection (Fig. 3c), resulting in slowed OA progression³⁶. HABP-functionalized PEG hydrogels implanted in rat osteochondral defects enhanced cartilage formation compared with unmodified hydrogels³⁷.

Chondrocytes. Chondrocyte affinity peptide (CAP; DWRVIIPPRPSA) enables direct targeting of chondrocytes³⁸. Identified via phage display, CAP leverages both positive charge and hydrophobicity to interact with amphiphilic membrane domains, exhibiting ~30-fold higher binding than the scrambled control peptide. CAP-modified polyethylenimine (PEI) complexes successfully transfected green fluorescence protein (GFP) plasmids in rat cartilage following IA injection (Fig. 3d (i)). Moreover, CAP-functionalized exosomes showed prolonged joint retention up to 6 days after IA injection to rat joints (Fig. 3d (ii)). Thus, CAP-functionalized exosomes enabled delivery of gene payloads such as MMP-13 siRNA³⁹ and miR-140⁴⁰ in rat OA models, resulting in enhanced cartilage protection and repair.

Synovium

The synovium is a soft-tissue membrane that structurally integrates joint components and plays a central role in maintaining joint homeostasis. Rich in immune cells, fibroblasts, and sensory nerves, it is a major contributor to arthritic inflammation (synovitis) and pain (Fig. 4a)⁴¹. The synovium closely interacts with cartilage, bone, and fat pad, significantly influencing disease progression⁴². Its high vascularity compensates for the avascular nature of cartilage, enabling nutrient exchange, waste clearance, immune regulation, and inflammatory signaling⁴³. Owing to this vascularity, the synovium not only represents the epicenter of inflammation in OA but also allows access to systemically delivered therapeutics. Moreover, its disease-specific phenotype, including macrophage infiltration, neovascularization, and altered matrix composition, offers opportunities to develop OA-specific targeting strategies that address both pain modulation and improved intra-articular drug retention⁴⁴ (Table 2).

Fig. 4. Synovium: tissue phenotype in OA and examples of targeting peptides.

a Schematic illustration of healthy versus OA synovium, showing key morphological and compositional changes as well as major molecular targets within the tissue. b Tissue specificity of the synovium-targeting peptide (HAP-1) and a scrambled control peptide (Con-P). Reprinted with permission from Elsevier, copyright 2003⁴⁵. c SyETP-modified nanoparticles exhibited preferential accumulation in inflamed rat joints (red box) compared with the non-targeting nanoparticles for up to 8 days following IV injection. Reprinted with permission from Elsevier, copyright 2018⁵⁰.

Table 2.

Synovium targeting peptides and their applications

Target	Peptide	Discovery	Application	Outcomes	Citation
Synovium Endothelium	CKSTHDRLC; Synovial endothelium targeting peptide (SyETP)	In vivo phage display	Peptide-small molecule conjugate: Cyclized SyETP-sinomenine	Selectively accumulated in the inflamed mice joints and alleviated acute inflammation	120
			Peptide-protein fusion: IL-4-MMP cleavage site-3x cyclized SyETP	Selectively targeted synovial graft and activated cargo protein in OA mice	51
			Polymeric-nanoparticle functionalization: SyETP-IgG2 on methotrexate loaded lactic acid/poly-caprolactone/PEG (tBNP)	Enhanced retention in inflamed mouse and rat joint, preventing neoangiogenesis	50
	DRL	Rational Design: analog of SyETP	Peptide-fluorophore conjugate: DRL-5-carboxytetramethylrhodamine (TAM)	Allowed confocal laser scanning microscopy (CLSM)-based screening on rat synovium-derived primary cells (RSPC)	121
FLS	SFHQFARATLAS; HAP-1	In vitro phage display	Peptide-peptide fusion: HAP-1-KLAK	Lead synovial apoptosis and concomitant reduction of OA rabbit’s inflammation and synovitis	45
			Carrier functionalization: HAP-1 on NF-κB-blocking peptide (NBD peptide)-loaded liposome	Selective targeting and inhibition of NF-kB pathway in arthritic rat	46
			Carrier functionalization: PEG microgel with HAP-1	Prolonged retention in inflamed rat joint	47
	HIQLSPFSQSWR; HAP-2	In vitro phage display	N/A	N/A
Synovium Neovasculature	CDCRGDCFC; RGD-4C	In vitro phage display	Peptide-peptide fusion: RGD-4C-KLAK	Reduced arthritis score of OA mice	52
Macrophage/ Immune cells	CRVLRSGSC; CRV	In vitro phage display	Peptide-small molecule conjugate: CRV-Dexamethasone	Improved therapeutic efficacy and safety of dexamethasone in OA mice	48
	CTTHWGFTLC	In vitro phage display	Carrier functionalization: surface modification of rapamycin-loaded mesoporous prussian blue nanoparticles (RAPA@MPB-MMP9)	Decreased macrophage pyroptosis, promoted mitophagy, and improved synovial lesions of OA mice	122
	cFLFLF	Rational Design: neutrophil formyl peptide receptor (FPR) antagonist	Peptide-contrast agent conjugate: cFLFLF-PEG-64Cu	Served as fluorescent probe for PET/CTa and showed enhanced contrast in OA mice	123

Synoviocytes. Fibroblast-like synoviocytes (FLS) are among the most extensively studied targets for peptide-mediated delivery. Phage display biopanning against the rabbit synoviocyte cell line HIG-82 identified the peptide SFHQFARATLAS (HAP-1), which exhibits high specificity for synovial cells (Fig. 4b)⁴⁵. A fusion construct of HAP-1 with the pro-apoptotic antimicrobial peptide KLAK induced selective synoviocyte apoptosis, and IA administration in a rabbit arthritis model significantly reduced inflammation and synovitis. HAP-1–modified microgels and liposomes also demonstrated improved joint retention and targeted delivery following IA injection in a rat arthritis model^46,47. In contrast, selective peptides for macrophage-like synoviocytes (MLS) and synovial macrophages remain underexplored. Most studies have therefore repurposed existing macrophage-targeting sequences for joint applications. For example, the macrophage-homing peptide CRV (CRVLRSGSC) has been utilized for reactive oxygen species (ROS)-responsive dexamethasone delivery via IV administration in a rat arthritis model⁴⁸. Although specific MLS markers have been identified, relatively few studies have explored the development of peptides uniquely selective for these synovial subpopulations⁴⁴.

Synovial endothelium. The synovial endothelium represents another promising therapeutic target. The Synovial Endothelium Targeting Peptide (SyETP; CKSTHDRLC), identified through combined ex vivo and in vivo phage display screening in a human/SCID mouse model, binds selectively to CD34⁺ synovial endothelial cells^49,50. Even though CD34 is broadly expressed across vascular and epithelial tissues, the exceptional joint specificity of SyETP suggests additional molecular interactions may underlie its targeting mechanism. SyETP is considered one of the most selective and well-validated synovial-targeting peptides⁵¹. In a rat rheumatoid arthritis model, poly-lactic acid, poly-caprolactone, and PEG nanoparticles with SyETP coating accumulated in inflamed joints following intraperitoneal administration with enhanced retention for at least 8 days compared to healthy or non-SyETP nanoparticle-treated groups (Fig. 4c)⁵⁰. Consequently, these synovium-targeting nanoparticles were able to deliver methotrexate and suppress synovial inflammation.

Synovial neovasculature. OA-associated neovascularization within the inflamed synovium also presents an attractive target for peptide-based interventions (Fig. 4a). The RGD-4C peptide, identified by phage display, exhibits high selectivity for the neovasculature through binding to αvβ3 and αvβ5 integrins⁵². IV injection of RGD-4C fused with the pro-apoptotic KLAK peptide induced selective destruction of synovial neovessels and significantly lowered arthritis severity scores in murine models. Such vascular-targeted strategies highlight the potential of leveraging pathological angiogenesis in OA for targeted drug delivery.

Subchondral bone (SB)

SB provides essential structural and metabolic support to cartilage. In OA, it undergoes aberrant remodeling characterized by increased bone turnover, hypervascularization, and aberrant sensory nerve ingrowth, collectively disrupting SB-cartilage crosstalk and accelerating disease progression (Fig. 5a)⁵³. As a result, SB is increasingly recognized as a key target for OA intervention. Targets such as mesenchymal stem cells (MSCs), osteoclasts, osteoblasts, and hydroxyapatite (HAP) of bone matrix are accessible via systemic administration, yet SB-directed strategies remain underexplored (Table 3). Most targeting peptides were developed for osteoporosis (OP) but held strong potential for OA therapy.

Fig. 5. Subchondral bone (SB): tissue phenotype in OA and examples of targeting peptides.

a Schematic illustration of healthy and OA SB, highlighting key structural and compositional alterations and major molecular targets within the tissue. b A fusion peptide combining a heptaglutamate (E7) sequence with a collagen-mimetic, osteoconductive DGEA motif enabled targeted accumulation in rat fibula up to 24 h following IV injection. Reprinted with permission from Elsevier, copyright 2013⁶³. c Dual modification with an osteoblast-targeting CH6 aptamer and a HAP-binding C11 peptide enabled enhanced bone accumulation of CH6-PAMAM-C11 dendrimers compared to unmodified PAMAM-PEG for up to 12 h after IV administration. Reprinted with permission from RCS, copyright 2021⁶⁴.

Table 3.

SB targeting peptides and their applications

Target	Peptide	Discovery	Application	Outcomes	Citation
MSC	EPLQLKM	In vitro phage display	Scaffold functionalization: Functionalized polycaprolactone (PCL) scaffold	Implanted scaffold improved endogenous MSCs mobilization to the defect sites of OA rat	55
Osteoblast	SDSSD	In vitro phage display	Carrier functionalization: Surface-modified polyurethane (PU) nanomicelles for anti-miR-214 delivery	Selective localization in bone and enhanced anti-miR-214 efficacy in alleviating OP progression in mouse	58
Osteoclast	TPLSYLKGLVTV	In vitro phage display	Carrier functionalization: Surface-modified red blood cell extracellular vesicles (RBCEVs) for anti-miR-214 delivery	Selective localization in bone and enhanced anti-miR-214 efficacy in alleviating OP progression in mouse	59
HAP	Poly(E)	Rational Design: inspired by repeating residues of hydroxyapatite-binding non-collagenous protein such as osteopontin and sialoprotein	Peptide-peptide fusion: Osteoinductive peptide (DGEA) fusion with heplaglutamate (E7)	Enabled retention of osteoinductive peptide in rat bone for up to 24 h	63
			Peptide-peptide fusion: Bone morphogenetic protein-2 (BMP-2)-derived peptide fusion with heptaglutamate	Facilitated retention of the BMP-2-derived peptide within allografts implanted in rat for at least one month.	124
	Poly(D)		Carrier functionalization: Surface-modified PLGA for simvastatin delivery	Targeted bone accumulation and improved simvastatin-induced bone growth in OP rat	125
	(DSS)6	Rational Design: inspired by repeating residues of dentin phosphoprotein	Carrier functionalization: Surface-modified DOTAP-based cationic liposomes for Plekho1 siRNA delivery	Selective Plekho1 siRNA enrichment in osteogenic cells enhanced bone formation in OP rat	126
	STDKTKREEVD; C11	Rational Design: derived for 11 residues at the C-terminus of amelogenin	Carrier functionalization: PAMAM dendrimer conjugated with C11 and osteoblast-binding CH6 aptamer (CH6-PAMAM-C11)	Rapid accumulation in rat bone 4 h post-injection with targeted accumulation at osteoblast sites	64

MSCs are attractive OA targets due to their self-renewal capacity and their ability to differentiate into both chondrocytes and osteoblasts, making them central to osteochondral repair⁵⁴. MSC-homing peptides can enhance cell recruitment and retention at defect sites. The phage-display-identified peptide EPLQLKM, when conjugated to polycaprolactone scaffolds, significantly increased MSC recruitment and reduced inflammatory cell infiltration in a rat cartilage defect model⁵⁵. Incorporation of matrix-binding motifs can further potentiate MSC-based therapies: hyaluronic acid- and chondroitin sulfate-binding peptide embedded within collagen-mimetic scaffolds promoted BM-MSC chondrogenesis⁵⁶, while MSCs coated with the collagen XII-binding peptide C5-24 and HA exhibited superior cartilage repair upon transplantation into OA rat joints compared to uncoated MSCs¹¹.

Osteoblasts and osteoclasts. Bone remodeling in OA is dysregulated due to the imbalance between osteoblast-driven formation and osteoclast-mediated resorption⁵⁷. The osteoblast-targeting peptide SDSSD, identified via phage display, was used to functionalize polyurethane nanomicelles and enabled targeted delivery of anti-miR-214. This enhanced osteoblast activity and improved bone structure after IV administration in a mouse osteoporosis (OP) model⁵⁸. For osteoclasts, the tartrate-resistant acid phosphatase (TRAP)-binding peptide (TPLSYLKGLVTV) was used to modify red blood cell-derived extracellular vesicles loaded with anti-miR-214, enabling targeted suppression of osteoclast activity and improved bone density in the same model⁵⁹.

Hydroxyapatite (HAP). Bone is primarily composed of HAP, an inorganic mineral that constitutes ~70% of its mass and confers structural integrity⁶⁰. Its abundance provides an opportunity for mineral-targeted delivery. HAP-binding peptides, typically enriched in acidic residues inspired by non-collagenous bone proteins, utilize electrostatic interactions with calcium. Such peptides have been exploited for osteoinduction, remineralization, and bone-targeted drug delivery⁶¹. For instance, FITC-labeled poly-aspartic acid demonstrated a femoral half-life of ~14 days after subcutaneous administration in mice⁶², while fusion of heptaglutamate with the osteoconductive DGEA motif achieved high accumulation in fibula within 24 h of IV injection in rats (Fig. 5b)⁶³. Using a similar approach, another HAP-binding peptide, C11 (STDKTKREEVD), was derived from amelogenin, a protein responsible for tooth enamel formation. Combined with osteoblast targeting aptamer CH6, CH6-PAMAM-C11 dendrimer exhibited a high accumulation in bones up to 12 h post-IV injection (Fig. 5c)⁶⁴.

Application: therapeutic peptide

Peptides can function as therapeutics by exerting anti-inflammatory, chondroprotective, regenerative, and analgesic effects—offering a multifaceted approach to disease modification and symptom relief (Table 4). Their growing therapeutic prominence is underscored by the broader pharmaceutical landscape, where 22 peptide drugs have received FDA approval within the past four years, more than half the total approvals from the previous decade^65,66.

Table 4.

OA therapeutic peptides

Category	Peptide	Discovery and Mechanism of Action	Modifications	Outcomes	Citation
Anti-inflammatory	Adrenocorticotropic hormone (ACTH) peptide; SYSMEHFRWGKPVGKKRRPVKVYPNGAEDESAEAFPLEF	Rational design: modified to improve half-life; binds to melanocortin 1 receptor	Fusion to latency affinity peptide to prevent inactivation in joint	Demonstrated anti-inflammatory OA attenuation effect in vivo	127,128
	Glucagon-like peptide 1 (GLP-1) agonist analogs; HAEGTFTSDVSSYLEGQAAKEFIAWLVKGRG	Rational design: modified to improve half-life; binds to GLP-1 receptor	Amino acid substitutions and side chain modifications	Activated P13K/Akt pathway in chondrocytes, protecting against IL-1 stimulation and ameliorating OA progression in vivo	69
	Melanocyte-stimulating hormone (MSH) peptides; α: SYSMEHFRWGKPV; β: AEKKDEGPYRMEHFRWGSPPKD; γ: YVMGHFRWDRFG	Rational design: modified to improve half-life; binds to melanocortin 1 receptor	Amino acid substitutions and fusion to latency affinity peptide	Modulated inflammatory activation of chondrocytes and synoviocytes in vitro and inhibited macrophage infiltration and cartilage degeneration in vivo	127–129
	Vasoactive Intestinal Peptide (VIP); HSDAVFTDNYTRLRKQMAVKKYLNSILN	Rational design: modified to improve half-life; binds to VPAC1 and VPAC2 receptors	Fusion to latency affinity peptide to prevent inactivation in joint	Alleviated cartilage damage through NF-κB pathway inhibition in vivo	127,130
Chondroprotective and regenerative peptides	C-type natriuretic peptide (CNP); GLSKGCFGLKLDRIGSMSGLGC	Rational design: modified to improve half-life; binds to NPRB receptor	Amino acid additions	Induced cartilage generation through activating cGMP/PKGII pathways	77,131
	FGFR1-binding peptide (R1-P1); GPPDWHWKAMTH	Phage display	N/A	Prevente cartilage degeneration and improved proteoglycan synthesis in vivo	78
	IL-1Ra-like peptide (DAP1-2); ADPQNRLELNQYNQWRDEQYVK	Rational design: 22 amino acids involved in binding between IL-1Ra and IL-1R1 selected	N/A	Reduced IL-1β after inflammatory stimulation in vitro and attenuated NF-κB signaling in vivo.	132,133
Pain relief peptides	Icatibant, DArg-Arg-Pro-Hyp-Gly-Thi-Ser-DTic-Oic-Arg	Rational design: analog of bradykinin	Amino acid substitution to non-canonical amino acids	Alleviated hyperalgesia in vivo after cytokine-induced arthritis, and restored weight bearing	82,83
	CGRP 8-37; VTHRLAGLLSRSGGVVKNNFVPTNVGSKAF	Rational design: analog of CGRP	Amino acid substitutions and side chain modifications	Locally blocked endogenous CGR reversed joint nociceptor responses in vivo	134

Anti-inflammatory peptides. Given OA’s strong inflammatory component, many therapeutic peptides have been designed to suppress proinflammatory signaling and halt disease progression. A promising candidate is glucagon-like peptide-1 (GLP-1), a clinically approved drug for diabetes and obesity that acts via the GLP-1 receptor (GLP-1R). Discovery of GLP-1R expression on chondrocytes revealed that GLP-1 protects against IL-1–induced inflammation by inhibiting the PI3K/Akt and NF-κB pathways^67,68. Beyond chondrocytes, GLP-1 suppresses inflammatory signaling in synovium, bone, and adipose tissue, underscoring its potential as a broad anti-inflammatory OA therapy^69,70. Current clinical trials primarily assess GLP-1’s indirect disease-modifying effects via weight loss, yet preclinical evidence indicates a direct joint-protective role, motivating further design of GLP-1–based peptides optimized for OA intervention.

Another notable anti-inflammatory agent is vasoactive intestinal peptide (VIP), a neuropeptide with broad immunomodulatory properties⁷¹. IA administration of VIP in preclinical OA models reduced synovitis and slowed disease progression by engaging G-protein–coupled receptors that suppress cytokine release and macrophage infiltration^72–75. However, the specific molecular role of VIP in cartilage homeostasis and OA pathophysiology remains incompletely defined, warranting deeper mechanistic study.

Chondroprotective and regenerative peptides. C-type natriuretic peptide (CNP), a hormone that regulates vascularization, angiogenesis, and cartilage growth during development, has established clinical relevance through its approval for achondroplasia treatment⁷⁶. In preclinical OA models, CNP treatment rescued cartilage by binding to natriuretic peptide receptor B (NPR2) on chondrocytes and activating the cGMP/PKGII pathway, which stimulates proteoglycan and collagen synthesis⁷⁷. Another chondroprotective candidate is R1-P1, a fibroblast growth factor receptor 1 (FGFR1)–binding peptide identified through phage display. FGFR1 activation drives cartilage degradation and suppresses ECM synthesis; accordingly, R1-P1 administration in rat OA models and human articular cartilage explants attenuated OA progression⁷⁸.

Analgesic peptides. While anti-inflammatory and regenerative peptides may indirectly reduce pain by limiting inflammation and restoring tissue homeostasis, many DMOADs have failed preclinical and clinical development because they do not provide lasting patient-relevant outcomes such as pain relief⁷⁹. This has spurred the development of analgesic peptides. Icatibant is a B2 receptor antagonist inspired by the natural nonapeptide of bradykinin^80,81. In a rat cytokine-induced OA model, icatibant alleviates knee hyperalgesia and restores weight-bearing capacity^82,83, though further validation in widely used injury-induced OA models is needed. Another approach involves calcitonin gene-related peptide (CGRP) receptor antagonists, derived from residues 8–37 of CGRP, a key mediator of mechanosensitivity and peripheral sensitization in OA joints. CGRP antagonists have demonstrated potential for pain reduction in preclinical synovial models^84,85.

Challenges and future direction

Increasing peptide stability against enzymatic degradation

A major limitation of peptide-based therapeutics is their inherently poor pharmacokinetics: low bioavailability, rapid degradation, and short systemic half-lives⁴. These challenges are further amplified in OA due to elevated protease activity and joint washout through SF clearance. To address these limitations, various strategies have been developed to enhance peptide stability against enzymatic degradation, including sequence optimization, bioconjugation, and structural stabilization (Fig. 6). Despite their importance, the pharmacokinetics of targeting peptides within joints remain poorly characterized, explaining the limited efforts to enhance their stability, whereas systemically delivered therapeutic peptides have been extensively optimized.

Fig. 6. Schematic illustration of strategies to improve peptide pharmacokinetics.

a Sequence optimization: optimization of peptide sequences by replacing residues at enzymatic cleavage sites enhances resistance to both serum enzymes (e.g., neutral endopeptidase (NEP) and dipeptidyl peptidase (DPP)) and tissue-specific enzymes such as matrix metalloproteinases (MMPs) and aggrecanases, thereby increasing stability. b Bioconjugation: attachment of antibodies, proteins, or non-protein partners such as immunoglobulins, transferrin, PEG, glycine, or fatty acids increases peptide size directly or indirectly via serum and synovial fluid protein binding. The larger size and steric shielding reduce renal clearance, protect against reticuloendothelial system (RES) uptake, and limit enzymatic degradation. c Structural stabilization: introduction or substitution of cysteine residues forms intramolecular disulfide bridges, generating a cyclic and rigid conformation that restricts access to cleavage sites, thereby enhancing stability.

Sequence optimization involves substituting residues at enzymatic cleavage sites, including those beyond the canonical amino acids (Fig. 6a). Unnatural amino acids expand this design space through stereochemical variants (e.g., D-isomers), backbone-modified chemistries (peptoids; β-, γ-, or ε-amino acids), and side-chain–engineered residues⁸⁶. The cleavage sites of native GLP-1 by neutral endopeptidase (NEP) and dipeptidyl peptidase IV (DPP-IV) are well defined, and clinically approved analogues incorporate substitutions that extend half-life⁸⁷. Inclusion of unnatural amino acids further broadens this modification landscape—such as Ala8 → 2-aminoisobutyric acid (Aib) in GLP-1 analogues, which prevents DPP-IV degradation⁸⁷, or incorporation of D-isomers in icatibant⁸⁸, which enhances receptor binding and proteolytic stability.

Bioconjugation can simultaneously mitigate enzymatic degradation, reduce renal clearance, and shield against reticuloendothelial uptake by linking peptides to larger partners at cleavage-prone sites (Fig. 6b). Conjugation to antibodies or proteins (e.g., immunoglobulins, transferrin) or non-protein partners (e.g., PEG, glycans, fatty acids) is commonly used to extend the systemic half-life of therapeutic peptides⁸⁹. For instance, clinically approved GLP-1 analogs are conjugated to immunoglobulins, PEG, or fatty acids, which extend their serum half-life up to 15 h^90,91. However, for intra-articular delivery, larger peptides may face steric hindrance within the dense joint extracellular matrix, limiting their transport.

Structural stabilization offers another route to enhance proteolytic resistance (Fig. 6c). Many peptides are intrinsically unstructured, making them susceptible to cleavage. CDPs, stabilized through intramolecular disulfide bridges, maintain a rigid conformation and stable surface charge—features that enhance both proteolytic resistance and cartilage targeting efficiency³⁰.

Incorporating in silico method for peptide discovery and optimization

In silico methods are gaining traction and potentially could be utilized to design and/or screen targeting and therapeutic peptides. Machine learning models such as AlphaFold2 and RoseTTAFold can predict protein structures with remarkable accuracy, and the continual expansion of computational power has positioned in silico approaches as an integral component in modern drug development. Additionally, molecular dynamics (MD) simulations also offer a powerful means to explore peptide conformational landscapes, with techniques such as replica exchange MD, metadynamics, and umbrella sampling.

In silico methods can be applied to both rational peptide design and large-scale screening. Zhang et al. used structure-guided modeling of ADAMTS4 with its endogenous inhibitor TIMP3 to identify a key inhibitory loop, which was optimized to yield a ninefold improvement in ADAMTS4 inhibition validated in vitro⁹². Similarly, large-scale in silico screening can markedly accelerate peptide discovery. For example, CDP variants previously explored for cartilage targeting were adapted for bispecific T-cell engagers. In this study, a library of 4,298 CDP peptide scaffolds was modeled using Rosetta. Low-energy computational docking and mammalian display screening filtered them down to 4 distinct scaffolds for in vitro validation. The top-performing scaffold was incorporated into a CD3-binding bispecific T-cell engager that successfully prolonged survival in a tumor-bearing mouse model⁹³. Such approaches can also model peptide-enzyme interactions to minimize protease cleavage or screen peptide libraries against OA-associated proteases to improve stability and specificity.

Study model weaknesses: In vitro model

Conventional evaluation of OA therapeutics, including peptide-based candidates, primarily relies on in vitro systems such as 2D/3D cell cultures or tissue explants. These approaches offer simplicity, accessibility, and compatibility with early-stage screening, enabling controlled experimentation in otherwise challenging environments^19,94,95. However, they fall short of recapitulating joint-level physiology, lacking key features such as inter-tissue crosstalk, synovial fluid dynamics, and physiologic mechanical loading. As a result, many therapeutic candidates advance prematurely to in vivo studies, highlighting the limitations of current in vitro models. Co-culture systems have improved physiological relevance by incorporating synovial cells or tissues, yet features such as spatial organization and biomechanical loading remain underrepresented. The growing emphasis on the 3 R principles (replacement, reduction, and refinement) further underscores the need to develop advanced, human-relevant in vitro systems⁹⁶.

The humanized joint-on-chip (JOC) platform provides a next-generation alternative, modularly reconstructing joint tissues within a microfluidic system⁹⁷. Each unit mimics native architecture and dynamics through integrated 3D cultures and controlled flow. Current JOC systems recapitulate diverse OA phenotypes, including injury, inflammation, and pain, by applying biomechanical loading and cytokine/chemokine cues^98–102. Although primary human chondrocytes, synovial fibroblasts, and immune cells from OA patients have been used, limited availability and donor variability remain challenges^99,103. A solution to these issues is to utilize induced pluripotent stem cells (iPSCs) as the source of the cells. Integration of induced iPSC-derived organoids within JOC systems could combine biological fidelity with precise environmental control, advancing physiologically relevant modeling of human OA.

A representative in vitro model based on humanized JOC offers a promising platform for advancing peptide-based targeted drug delivery systems and therapeutic peptide development. While complete replacement of animal models remains a long-term goal, the humanized JOC can help bridge this gap. Its multi-tissue composition and inclusion of joint-relevant dynamics and loading can enable evaluation of targeted delivery features not captured previously by conventional in vitro systems. However, the lack of standardization, variability in experimental protocols, technical complexity, and high costs pose significant challenges that must be addressed before humanized JOC systems can be widely adopted.

Unexplored target in the joint

Most peptide research has focused on cartilage and synovium, yet OA is a whole-joint disease involving multiple tissues. There remains limited exploration of peptides targeting other joint tissues, which could open new avenues for OA therapeutics. SB plays a key role in mechanical support, nutrient exchange, and signaling. While peptide-based SB targeting has been explored for OP treatment and imaging, its application to OA remains nascent. Harnessing the vascularized nature of SB could enable peptide-assisted systemic delivery to osteoblasts, osteoclasts, and MSCs.

SF maintains joint homeostasis by providing lubrication and nutrient transport. It contains potential targets such as HA, serum proteins, proteases, and polysaccharides. Electrostatic interactions with HA have been utilized to entrap drug conjugates in SF, suggesting similar potential for enhancing peptide residence time^104,105. HABPs, previously used for cartilage and bone targeting, may be adapted for this purpose. However, overly strong interactions may compromise delivery efficiency by restricting tissue diffusion or masking functional ligands^18,106.

Other joint tissues, such as the ligament, meniscus, and infrapatellar fat pad (IFP), also merit exploration. The ligament and meniscus are dense, collagen I–rich connective tissues with low cellularity^107,108. A decorin-derived peptide, LRELHLNNN, exhibits strong binding to fibrous tissues through collagen I interactions and may serve as a targeting motif for these structures¹⁰⁹. IFP, a vascularized and innervated adipose tissue, contributes to OA inflammation through adipocyte-immune signaling and serves as a reservoir for reparative MSCs¹¹⁰. Adipocyte-targeting peptides such as CKGGRAKDC, which bind prohibitin on vascular adipocytes and can deliver pro-apoptotic cargo to white adipose tissue, exemplify potential for IFP-directed OA therapy¹¹¹. Despite such promise, strategies targeting these tissues remain scarce, partly due to their limited involvement or ambiguous roles in OA pathogenesis.

Conclusion

Peptides represent a versatile and modular platform for OA therapy, functioning both as targeting ligands and as therapeutic agents with anti-inflammatory, regenerative, or analgesic effects. Their clinical translation, however, is constrained by poor stability and rapid clearance within the joint. Significant efforts have been made to address these stability challenges, aiming to enhance peptide durability for targeted delivery to the joint. Looking forward, advances in in silico design, humanized joint-on-chip systems, and the exploration of underutilized joint targets will accelerate discovery and optimization. Together, these developments highlight a future where peptide-based approaches enable precise, stable, and multi-tissue OA therapies with true disease-modifying potential.

Author contributions

B.H. and A.B. conceived the idea. B.H., H.Z., and A.S. conducted the literature review and curated the data. B.H. prepared Figs. 1–3, 5 and Tables 1 and 3. H.Z. prepared Fig. 4 and Table 2. A.S. prepared Fig. 6 and Table 4. All authors contributed to conceptualization and participated in writing, reviewing, and editing the manuscript.

Data availability

No datasets were generated or analysed during the current study.

Competing interests

The authors declare no competing interests.