Overcoming Barriers for Intra-articular Delivery of Disease-Modifying Osteoarthritis Drugs

Abstract

Despite four decades of research in intra-articular drug delivery systems (DDS) and two decades of advances in disease-modifying osteoarthritis drugs (DMOADs), there is still no clinically available disease modifying therapy for osteoarthritis (OA). Multiple barriers compromise intra-articular DMOAD delivery. Although multiple exciting approaches have been developed to overcome these barriers, there are still outstanding questions. We make several recommendations that can help in fully overcoming these barriers. Considering OA heterogeneity, we also propose a patient-centered, bottom-up workflow to guide preclinical development of DDS-based intra-articular DMOAD therapies. Overall, we expect this review to inspire paradigm-shifting innovations for developing next generation DDS that can enable clinical translation of intra-articular DMOADs.

Towards Intra-articular Disease-Modifying Therapies

Osteoarthritis (OA) is a complex and multifactorial disease that impacts the whole joint, resulting in pain, tissue damage and eventually loss of function. Pathological changes in OA joints are mainly characterized by cartilage breakdown, subchondral bone remodeling, osteophyte formation, and synovial inflammation. Additionally, changes in muscles, nerves and local fat pads may also contribute to OA progression. The structural changes are also accompanied by dysregulation of complex metabolic and biochemical pathways, such as Wnt signaling and matrix degrading enzymes, leading to disturbed joint homeostasis. Over time, the disease continues to progress, and eventually knee replacement surgery is required.

Emerging advances in the discovery of disease-modifying osteoarthritis drugs (DMOADs) (see glossary) have brought hope for OA patients [1]. As of now, about 40 ongoing clinical trials are actively evaluating DMOADs [2]. Unfortunately, previous clinical trials for DMOADs have experienced a high failure rate and to date, no DMOAD has been approved for clinical use. Therefore, it is not surprising that the Food and Drug Administration (FDA) has recognized OA as a serious disease with an unmet medical need for disease-modifying therapies [3].

A major obstacle towards clinical translation of DMOADs is their insufficient accumulation in OA joints upon systemic administration, which results in low therapeutic efficacy and off-target effects [4,5]. Intra-articular injection can maximize local concentration of DMOADs in joints; however, DMOADs exhibit short half-lives of around 0.1–6 h in joints [4]. Intra-articular drug delivery systems (DDS) can increase the joint residence time of DMOADs by enabling sustained drug release [6]. DDS and can also be engineered to target specific tissues in joints and to cross physiological barriers such as dense cartilage matrix [7]. Over 3200 studies published in the last 20 years have focused on developing intra-articular DDS for a wide range of joint diseases; however, clinical translation of intra-articular DDS for OA therapy has shown limited success with only one FDA-approved product (Zilretta®) to date [8], which is intended for pain relief. A clear, unmet need therefore exists to step back and rethink the strategies for design and development of intra-articular DDS for DMOADs.

In this review, we first summarize major advances in the field of intra-articular DDS for OA therapy, including the current clinical landscape. Next, we discuss barriers towards intra-articular delivery of DMOADs, describe the state of the art approaches to address those barriers, and highlight the outstanding questions for future research. We then present a forward-looking “bottom-up” workflow to guide pre-clinical development of intra-articular DDS for disease-modifying therapies. Finally, we briefly discuss critical considerations for clinical translation, including scale-up and manufacturing, and clinical trial design. Overall, we expect this review to serve as a guiding document to inspire innovations in developing next-generation intra-articular DDS that can enable successful clinical translation of DMOADs.

Current Status of Intra-articular DDS

Last four decades have witnessed numerous pre-clinical and clinical efforts towards the development of intra-articular DDS. Figure 1 provides a comprehensive timeline of the seminal discoveries and clinical advancements. The first use of intra-articular injection in arthritic joints dates back to 1951, when Hollander and coworkers administered hydrocortisone acetate for treating joint pain in patients [9]. To improve joint residence time of drugs, multiple biodegradable and biocompatible DDS have been explored. In 1978, Dingle and coworkers reported a liposome based formulation of cortisol to treat acute experimental arthritis in rabbits via intra-articular injection [10]. Later, in 1987, Stimpel and Weder found that increasing the size of liposomes from 160 to 750 nm improved the joint retention of encapsulated drugs by 2.6 times [11]. It was only recently in 2016 when a liposomal formulation (TLC-599) for intra-articular therapy of OA completed a phase II clinical trial and advanced to phase III in 2019 (Table 1) [12,13]. TLC-599 consists of dexamethasone loaded liposomes, designed to provide sustained pain management in OA. To date, Lipotalon® is the only liposomal product available in market for intra-articular injection of dexamethasone palmitate, but only in Germany [14].

Figure 1. Timeline showing seminal discoveries and clinical advancements in the field of intra-articular drug delivery.

Paramount events as early as 1951 have paved the way for the current status of intra-articular drug delivery for OA therapy. However, there has been an exponential increase in the research pace over past 10 years. Specifically, multiple intra-articular therapeutics based on different types of DDS (TLC599, FX005, FX006, Cingal, SI-613, and Condrotide Plus) entered into clinical trials, with FX006 (triamcinolone acetonide loaded PLGA microparticles) gaining FDA approval in 2017. However, clinical translation of intra-articular DDS for DMOAD delivery is still underwhelming, with only one formulation (Condrotide Plus) currently in clinical trial. The timeline shows key pre-clinical advancements in intra-articular DDS for OA therapy (Yellow boxes), FDA approval or initiation of clinical trials for intra-articular DDS based therapeutics for OA (Red boxes), and key pre-clinical developments related to DDS based intra-articular DMOAD therapies (Green boxes). Abbreviations: IA = intra-articular; HA = hyaluronic acid; DDS = drug delivery system; PLGA = poly(lactic-co-glycolic) acid; FGF = fibroblast growth factor; PNT = polynucleotide; NP = nanoparticle; TH = Triamcinolone Hexacetonide; MP = microspheres; LNP = lipid nanoparticles; Col II = Type II Collagen, MMP = Matrix Metalloproteinase.

Table 1.

Summary of clinical trials focused on intra-articular drug delivery for osteoarthritis¹²

Product	Drug	Delivery System	Phase	Status	Safety/Efficacy/Pain Relief Measures	Trial # Reference
TLC599	DSP3	Multi-layer lipid membranes	3	Active	Efficacy: TLC599 showed significant WOMAC-pain reduction from baseline vs. placebo from Week 1 to 24	NCT02803307 NCT03005873
EP-104IAR	FP4	Poly(vinyl alcohol) coating	1	Completed	Safety: Minimal vacuolation observed in low-dose group while mild vacuolation was observed in high-dose group on Day 7.	NCT02609126
Cingal	TH5	Hydrogel	3	Recruiting	Efficacy: Cingal had significantly lower WOMAC Pain scores than the saline group through 26 weeks; also superior to Monovisc for pain relief at 1 and 3 weeks	NCT04231318
SI-6134	DF6	Hydrogel	3	Completed	Efficacy: WOMAC and 50-foot walk test pain scores from baseline change for SI-613 was significantly higher than placebo group over 12 weeks; consistent with the prolonged improvement in 11-point NRS score	NCT03209362 UMIN000015858 JapicCTI-183855
sc-rAAV2.5IL-1Ra	IL-1Ra7	scAAV8	1	Recruiting	Efficacy: Median Mankin score in the high-dose vector group was significantly lower than that in the control group at the Day 7 and 180; However, therapeutic effects of the vector were not observed on Day 364	NCT02790723
FX-201	IL-1Ra	HDAd vector9	1	Recruiting	Efficacy: Pilot study – 2 out of 5 patients treated with FX201 in the low-dose, single ascending dose cohort reported substantial improvement in pain out to Week 24	NCT04119687
Condrotide Plus	PNT10	Hydrogel & Mannitol	N/A	Completed	Efficacy: Condrotide Plus showed better results than hydrogel alone in improving KSS total score and pain category. A significant reduction in the WOMAC score was observed over time for both groups.	NCT02417610
FX006	TA11	PLGA polymer12	2	Completed	Safety: About 33.3% adverse event rate for shoulder injection and 46.7% for hip injection	NCT02357459

¹For products with multiple clinical trials, the information shown in Table 1 reflects the most updated trial information.

²If clinical results are not available, we replaced them with pre-clinical animal results indicated by * sign

³DSP: dexamethasone sodium phosphate;

⁴FP: fluticasone propionate;

⁵TH: triamcinolone hexacetonide;

⁶DF: diclofenac;

⁷IL-1Ra: interleukin-1 receptor antagonists;

⁸scAAV: self-complementary adenovirus vector;

⁹HDAd: helper-dependent adenovirus;

¹⁰PNT: polynucleotides;

¹¹TCA: triamcinolone acetonide;

¹²PLGA polymer matrix used for FX006 is 75:25 in composition and the drug loading is 25% (w/w)

In 1984, Hunneyball and coworkers performed the first study to compare biocompatibility of different polymeric and protein-based materials in rabbit knee joints [15]. In 1998, FDA approval of Lupron Depot®, the first poly(lactic-co-glycolic) acid (PLGA) copolymer based injectable DDS for cancer, put a spotlight on PLGA as a potential DDS for joint diseases as well. The first attempt of using PLGA as an intra-articular DDS for OA was reported shortly after in 2000, where Hincal and colleagues prepared PLGA particles loaded with Naproxen Sodium, which exhibited statistically significant difference in arthritic alleviation compared to vehicle control [16]. In 2010, polymeric particles entered clinical trials as an intra-articular DDS. FX005, a PLGA microsphere-based formulation of p38/MAPK inhibitor was developed by Flexion Therapeutics and AstraZeneca [17]. However, clinical studies on FX005 were later discontinued when FX006, a PLGA microsphere encapsulating TA gained FDA approval in 2017 [18]. To date, this remains the only FDA-approved DDS-based intra-articular treatment for OA.

In 2008, Rothenfluh et al made the first attempt to achieve targeted delivery of nanoparticles to cartilage upon intra-articular injection. Using poly(propylene sulfide) nanoparticles functionalized with a collagen II-binding phage-display peptide (WYRGRL), the group reported a 72-fold increase in cartilage targeting compared to non-functionalized nanoparticles [19]. Following this success, a myriad of targeting strategies blossomed. In 2014, Grodzinsky and colleagues became the first group that demonstrated electrostatic interaction-based passive targeting of cartilage and showed that positively charged avidin molecules had 400-fold higher uptake in cartilage explants than their neutral counterparts [20]. One year later, Zhang et al discovered another peptide (DWRVIIPPRPSAC) with cartilage homing capability and functionalized it on polyethyleneimine particles to deliver anti-HIF-2a siRNA [21]. This was also the first report showing siRNA delivery into OA joints using nanoparticles. Most recently in 2021, Bedingfield et al. utilized collagen II-binding antibody functionalized polymeric nanoparticles for intra-articular delivery of matrix metalloproteinase (MMP) 13 siRNA and achieved prolonged retention in joint and OA attenuation [22].

Hydrogel-based DDS have also been investigated for intra-articular drug delivery. The seminal study that inspired the development of hydrogel-based intra-articular DDS dates back to 1971 when Rydell and Balaz injected hyaluronic acid (HA) into healthy and OA knees of dogs and owl monkeys and observed an elastic cushioning effect [23]. Later, in 1982, Namiki and coworkers performed intra-articular injection of high molecular weight HA in OA patients with low to moderate grade osteoarthritis. HA showed pain-relieving effects in 71% of patients [24]. HA hydrogel was FDA-approved in 1997 as an intra-articular viscosupplement for symptomatic pain relief [25], which greatly promoted the development of HA-based hydrogel formulations for delivery of a variety of agents, including small molecule drugs, biologics and even cells [6]. In 2006, Inoue et al. made the first ever attempt to deliver a DMOAD - basic fibroblast growth factor (bFGF) using an intra-articular DDS [26]. bFGF was loaded into gelatin hydrospheres and showed sustained release kinetics and OA suppression in rabbits. Compared to just a few particle-based intra-articular DDS, a larger number of hydrogel-based intra-articular DDS advanced to clinical trials. Cingal, a crosslinked HA gel loaded with triamcinolone hexacetonide (TH), entered clinical trial in 2013 and is currently in phase III clinical trials [27]. SI-613, an HA hydrogel-based formulation of diclofenac developed by Seikagaku Co. showed significant improvement in the Western Ontario and McMaster Universities Arthritis Index (WOMAC) score compared to placebo in a phase III study [28,29]. Interestingly, a polynucleotide/HA injectable gel system (Condrotide Plus) is the only DMOAD loaded DDS that advanced into clinical studies in 2014, and is currently being investigated for joint tissue regeneration in OA [30].

Overall, research related to intra-articular drug delivery has increased exponentially. A plethora of intra-articular DDS with varying size and material composition has been developed for the delivery of a wide range of therapeutics. A detailed summary of these DDS has been systematically compiled in multiple reviews published recently [31,32]. However, clinical translation of intra-articular DDS for OA therapy has been underwhelming. A clinicaltrials.gov search shows that fewer than 20 clinical trials involving device or DDS assisted drug delivery for OA are under active investigation, and only 1 clinical product (FX006) is FDA approved to date (Table1). Clinical translation of intra-articular DDS is partly limited due to multiple barriers that compromise the local pharmacokinetics and biodistribution of drugs in the injected joint. In the following section, we discuss these barriers in detail and highlight gaps in approaches that have been developed to overcome these barriers.

Barriers for DMOAD Delivery in Joints

Although native anatomical components in a joint provide protection, stability, and mobility, they also result in multiple barriers that impede DMOADs and DDS from reaching their targets. The following section provides a detailed description of these barriers (Figure 2), state of the art approaches to overcome them, and outstanding questions for future research.

Figure 2. Barriers towards intra-articular delivery of DMOADs.

While native structures, including cartilage, synovium and menisci provide protection, stability, and mobility to a joint, they also result in multiple barriers that impede DDS and DMOADs from reaching their target. We have classified these barriers into four categories. 1) Joint clearance: For particle-based DDS, clearance is primarily driven by their size, which defines particle transport through the synovium membrane, their uptake by immune cells, and efflux through lymphatics and microvasculature. For hydrogel-based DDS, clearance is primarily governed by material degradation; 2) Repeated mechanical loading: Repeated mechanical loading of joints due to strenuous activities such as running, exercise and playing sports can impact the structural integrity of DDS, causing rapid drug release; 3) Dense cartilage matrix: Cartilage is the most important target for DMOAD; however, its dense matrix poses a huge barrier to the penetration of DDS. This issue gets further aggravated by the fact that the majority of chondrocytes reside in the middle and deep zones of the cartilage. Therefore, to reach chondrocytes, drug carriers are required to penetrate through the densely packed and negatively charged collagen fibril network; and 4) Non-specific biodistribution: DMOADs are designed to inhibit or activate a specific biological pathway and should therefore target the right cells and tissues to achieve maximum efficacy. Non-specific biodistribution of DDS within joint can result in compromised efficacy and off-target side effects, especially due to repeated administration over a long term.

Joint Clearance

Clinical translation of promising DMOADs has remained challenging due to their rapid clearance after intra-articular administration [33]. Chemical modification of drug molecules to increase their hydrophobicity has been attempted to delay drug elimination. For instance, methylprednisolone acetate (MPA), a prodrug of methylprednisolone (MP) showed 8-day joint retention as opposed to 24-hour retention shown by unmodified MP [34,35]. Another approach that delays joint clearance of drugs involves formulating them with amphipathic excipients such as polysorbate 80 (PS80). Kenalog-40, an FDA approved formulation of TA suspended in carboxymethylcellulose sodium and PS80, demonstrates efficacy for several weeks with a single intra-articular injection [36]. Although no direct evidence is available, it is likely that such an extended duration of effect is due to PS80, an amphipathic surfactant [37]. However, it might be possible that additional mechanisms also contribute towards prolonged effect of Kenalog-40. Elucidating these mechanisms could open new avenues to prolong the joint residence time of DMOADs.

The most explored strategy for extending joint residence time of DMOADs involves their encapsulation in rationally engineered DDS. DDS that are too large to cross the ‘leaky’ synovium and to escape the trans-synovial flux in joint [14] offer a great opportunity to extend joint residence time of therapeutics. Previous studies have shown that particles with size less than 300 nm escape freely from the joint cavity via lymphatics and microvasculature regardless of the joint inflammation status [38]. Three μm-microparticles can leak through the inflamed synovial membrane but can be retained in the synovium. Ten μm or larger microparticles on the other hand are mostly retained in the synovial cavity, independently of the joint inflammatory status [38]. Advantage of >10 μm sized particles for long-term intra-articular drug delivery is also clearly evident from Zilretta®, which consists of 42 μm sized PLGA particles loaded with TA and demonstrates efficacy for about 12-weeks with a single injection [39].

Phagocytosis of particles by immune cells in joint also contributes towards their clearance. It has been shown that particles <5 μm in diameter can be eliminated via phagocytosis by synovial resident and recruited macrophages, which could also lead to an inflammatory response [40–42]. Additionally, particles may also be phagocytosed by dendritic cells (DCs) in synovium, which could trigger the migration of activated (DCs) to lymph nodes to activate T cells [43]. Particles may also experience surface adsorption of complement proteins in synovial fluid, resulting in their clearance by mast cells via opsonin-dependent phagocytosis [6]. Incorporating neutralizing cationic coatings on nanoparticles [44], and reducing non-specific hydrophobic and hydrophilic interactions of particles with proteins in synovial fluid have shown to reduce the uptake of particles by immune cells [44]. Recently, cell-mimicking nanoparticles have also attracted attention, wherein particles cloaked with an endogenous cell membrane inherit the “self” identity from the source cells and escape immune activation [6].

Clearance of particles from joints also depends upon their porosity and degradation kinetics, which has not been studied extensively in previous studies. Therefore, we argue that while designing particle-based DDS for DMOAD delivery, porosity and degradation kinetics should be given due consideration to maximize joint residence time.

For hydrogel-based DDS, the most critical factor governing clearance is their degradation in joint environment. Therefore, extensive in vivo studies should be performed to evaluate the kinetics of hydrogel degradation in joints. These studies should ideally be performed in arthritic animals, especially if the material is susceptible to degradation by bio-chemical molecules, including certain enzymes, pH, or peroxides that are overexpressed in OA. Materials that are rapidly degraded in joint environment can be chemically tuned to slow down their degradation. For example, HA is rapidly degraded by hyaluronidase enzyme, present in both healthy and OA joints [45]. Therefore, for intra-articular drug delivery, HA is usually cross-linked to reduce its enzymatic degradation [46]. Although chemical cross-linking is an excellent tool to tune hydrogel degradation, it can also complicate scale-up and manufacturing of the DDS. Additionally, certain cross-linking agents can be toxic to the joint microenvironment and can compromise the biocompatibility of DDS. Therefore, to minimize scale-up and toxicity concerns, considerations must be given to the choice of cross-linking agent and complexity of the chemical reaction. Different cross-linking techniques for hydrogel preparation along with their pros and cons have been previously reviewed elsewhere [47,48].

An unexplored area is the interaction of hydrogel with joint macrophages (See Outstanding Questions). Although tissue resident macrophages have proven to phagocytose particles [49], their relationship with hydrogel materials such as HA is more complicated. While high molecular-weight HA is reported to inhibit macrophage phagocytosis and prevent release of reactive oxygen species [50], fibers and fragments of HA were shown to induce macrophage activation and inflammatory response [51]. Therefore, future studies should be performed to understand similar relationships between other hydrogel materials and joint macrophages.

Outstanding questions.

• How do hydrogel-based DDS interact with joint-resident macrophages and how do such interactions contribute to the degradation and clearance of the DDS?

• How does the degradation kinetics of a DDS change with OA severity?

• Since particle size that is appropriate for cartilage penetration is also susceptible to rapid clearance via lymphatics and microvasculature, what new design modifications can be made to minimize rapid clearance of cartilage targeting particles?

• How does protein corona formation on cartilage targeting particles within the synovial fluid affect their clearance or mask their targeting moieties?

• How do changes in mechanical loading affect the joint microenvironment and what are its implications on the degradation or release kinetics of DDS?

• How local release kinetics of a DMOAD in joint impacts it’s therapeutic efficacy?

• How can OA patient classifications be performed with more precision, using novel technologies such as multi-omics and artificial intelligence?

It is also important to note that the clearance rate of injected DDS for DMOAD delivery might be affected by the extent of OA pathogenesis. OA progression induces changes in structure and composition of joint tissues, including synovium or cartilage [52]. These transformations can very likely affect the fate of DDS after intra-articular injection [53,54]. Therefore, we suggest that deep understanding of how material degradation kinetics changes with OA severity is necessary to accurately determine the joint residence time of DDS (See Outstanding Questions).

Non-Specific Biodistribution of DDS

A joint consists of multiple tissues, including, bone, cartilage, synovium, ligaments and meniscus, and each one of these is involved in one or more biological pathways that contribute to OA progression [52]. Typically, DMOADs are designed to inhibit or activate a specific biological pathway and should therefore specifically target the right tissue (Figure 3) to achieve maximum efficacy. Drug delivery systems that can target specific cells or tissues within joints can substantially enhance the therapeutic outcome of a DMOAD and minimize off-target effects [6].

Figure 3. Classification of DMOADs.

DMOADs are designed to inhibit or activate a specific biological pathway and should therefore specifically target the right tissue to achieve maximum efficacy. Broadly, DMOADs can be classified into four categories: 1) DMOADs that slow down or reverse cartilage degradation; 2) DMOADs that reduce synovium inflammation; 3) DMOADs that target subchondral bone pathophysiology; and 4) DMOADs that inhibit adipokines.

For DDS that are intended to localize in synovium, synovial macrophages are an excellent target due to their elevated population and inherent phagocytic capacity [39]. Passive targeting of synovium can be achieved by using particles with size between 0.3–10 μm [38], as they would exhibit maximum transport through the synovium membrane but minimal clearance through lymphatics and microvasculature [55]. To achieve passive targeting of cartilage, multiple reports have utilized cationic materials, which can electrostatically bind to negatively changed glycosaminoglycans [56]. Size is also a critical factor that impacts cartilage targeting and penetration. A few ex vivo studies have shown that materials up to ~55 nm can penetrate the full thickness of bovine cartilage [20,57]. However, there are also reports showing that particles with size around 6 nm can only penetrate the superficial layer of the cartilage [58]. These discrepancies might be due to differences in the cartilage damage across different studies. More systematic studies are therefore required to correlate penetration of particles with the extent of cartilage damage.

Although electrostatic interaction based passive targeting of cartilage is an interesting approach, it also has a few limitations. Firstly, positive charge on particles might get neutralized by anionic molecules in synovial fluid, including proteoglycan 4, hyaluronan and surface-active phospholipids [59]. Secondly, due to proteoglycan loss during OA, the effective charge density on cartilage might become less negative compared to healthy cartilage. Finally, particle size that is appropriate for cartilage penetration is also susceptible to rapid clearance via lymphatics and microvasculature (See Outstanding Questions). These limitations must be carefully considered while developing DDS that target cartilage via size and charge-based interactions.

Active targeting mechanisms that involve binding of the DDS to certain uniquely or overexpressed ligands and receptors on the target cell could offer more specificity for DMOAD delivery. WYRGRL, a six amino acid peptide that targets collagen-IIα1 in cartilage matrix, was conjugated to nanoparticles and demonstrated 71-fold higher accumulation in cartilage compared to non-targeted nanoparticles [19]. Other targeting ligands, including HA, certain peptides, and antibodies for collagen-II have also shown great potential in pre-clinical models [22,55,60,61]. Nevertheless, there are also a few underlying concerns with active targeting strategies. Firstly, decoration of ligands on the DDS could change their original physiochemical properties, such as size and surface charge, which might impact their clearance. Secondly, proteins present in synovial fluid might get adsorbed on the DDS resulting in a ‘protein corona effect’ thereby masking the active targeting ability. Lastly, expression of the target protein or receptor might vary with the disease stage, which can impact the targeting capability of the DDS. To understand these issues, there is already some work underway. For example, Allen and coworkers demonstrated that protein corona formation on microparticles injected into rat knee varies based on the stage of OA progression [62]. Later Stubelius and coworkers found that nanoparticles with longer poly(ethylene glycol) (PEG) chains exhibit reduced corona formation and are taken up more by the cartilage cells than nanoparticles with shorter or no PEG coating [63]. Overall, to develop efficient active targeting approaches for intra-articular DDS, more efforts are required to further enhance our understanding of DDS-protein interactions in joint. (see Outstanding Questions).

Another anatomical structure that could benefit from targeted intra-articular drug delivery is meniscus. Meniscal injuries often occur as tears in the avascular inner zone, which is incapable of regeneration [43]. Meniscal regeneration has been mainly attempted using implantable scaffolds loaded with growth factors and mesenchymal stem cells [64–66]. However, compared to implantable scaffolds, intra-articular delivery of agents that can regenerate meniscus is more patient compliant and less complex from a regulatory perspective. In a study by Kawanishi et al, intra-articular injection of synthetic microRNA-210 accelerated avascular meniscal healing in rats [67]. Delivery of such promising regenerative agents in menisci targeting DDS could further enhance their therapeutic efficacy. Therefore, we recommend that efforts should be directed towards developing strategies for targeted intra-articular delivery of regenerative agents to damaged meniscus.

Dense Cartilage Matrix

Cartilage is the most important target for DMOADs [68,69]; however, its dense matrix poses a huge barrier to the penetration of DDS [56]. This issue gets further aggravated by the fact that the majority of chondrocytes reside in the middle and deep zones of cartilage. Therefore, drug carriers have to penetrate through densely packed and negatively charged collagen fibril network. Bajpayee and Grodzinsky suggested a passive approach for cartilage penetration [56]. They proposed that weaker reversible charge-based interactions between cationic DDS and anionic proteoglycans in cartilage are key for DDS penetration. Based on this concept, avidin with a net charge of +20 mV was shown to penetrate a 1000 μm bovine cartilage explant ex vivo within 24 hours. Geiger et al also developed a PEGylated polyamidoamine (PAMAM) dendrimer-based cationic nanocarrier to deliver IGF-1 and identified that Gen 6 PAMAM dendrimer (58 kDa) with 45% of PEG surface groups achieved the most optimal penetration into bovine cartilage and prevented cartilage degeneration by 60% in rats [70]. Although the results are exciting, this study did not determine the percentage of injected nanocarriers that penetrated into the cartilage. This information is critical to determine the effective dosage that results in maximum therapeutic effect with minimum systemic toxicity. Also, efficacy of cartilage targeting nanocarriers was not compared to non-targeting nanocarriers, which makes it difficult to understand the true potential of this passive penetration approach.

Active approaches have also been developed for cartilage penetration. For example, WYRGRL modification of ~30 nm nanoparticles resulted in whole depth penetration of mouse cartilage (~50 um) [19]. Although both active and passive penetration approaches are promising, concerns related to cartilage targeting, as described in the previous section are also valid for cartilage penetration and must be addressed to unleash the full potential of cartilage penetrating DDS. One additional concern is that previous studies evaluating cartilage penetration have considered cartilage as a single entity and have not determined variabilities in partitioning of the DDS in different quadrants of the articular cartilage [22]. Anatomically, articular cartilage is divided into four quadrants. It exists on both the femoral condyle and tibial plateau sides and extends in medial and lateral directions on each side [71]. Since degeneration can happen in any of the four quadrants of the cartilage, we believe that it is critical to evaluate targeting and penetration of DDS to each quadrant of the cartilage.

Repeated Mechanical Loading

Since most DMOADs are intended for early phase of the disease [1], when patients are young and physically active, it is critical to consider the impact of mechanical loading of joints on the DDS and on the release kinetics of encapsulated DMOAD. None of the previously developed DDS for DMOAD delivery have been evaluated in physically active joints or have considered the impact of mechanical loading due to physical activity on the DDS and on the DMOAD release. Repeated mechanical loading of joints due to strenuous activities such as running, exercising and playing sports can be detrimental for the delivery system and could cause rapid drug release, thereby reducing the joint residence time and efficacy of the DMOAD [72,73]. Therefore, we propose that development of intra-articular DDS for DMOAD delivery should consider evaluating the impact of repeated mechanical loading on the DDS properties, including its microstructure, morphology, degradation rate and release kinetics. Korin et al demonstrated that high shear stress caused by vascular narrowing could break up PLGA based microscale aggregates into nanoparticles and release the encapsulated drug [74]. Xiong et al evaluated the stability of bFGF loaded PLGA microspheres under mechanical stress in vitro and found that microsphere degradation and bFGF release were faster under 4.0 MPa stress compared to non-stressed microspheres [75]. Although these studies were performed for non-OA applications, they clearly suggest that mechanical loading can impact both the structural and functional properties of DDS. It is also important to understand how repeated mechanical loading of joints impact the underlying physiological environment and OA pathophysiology, which in turn can also affect the degradation and release kinetics of DDS (see Outstanding Questions). The rationale for this stems from previous studies which have shown that acute or chronic high-intensity loads can directly damage cartilage extracellular matrix and shift the balance in chondrocytes from anabolic to catabolic activity [76,77]. It has also been shown that reduced mechanical loading due to joint immobilization creates catabolic responses within cartilage and cause cartilage thinning and tissue softening [78]. These processes are also accompanied with biochemical and metabolic changes, such as accumulation of degradative enzymes (e.g. MMPs, ADAMTs), cytokines (e.g. tumor necrosis factor-α, interleukin-1ß), and varied pH [79]. Such microenvironment changes can very likely influence the degradation or release kinetics of the DDS [80]. Therefore, we believe that gaining more insights into the impact of joint biomechanics on OA pathophysiology and joint microenvironment will provide invaluable inputs to engineer intra-articular DDS that can withstand repeated mechanical loading of joints.

Workflow for Designing Intra-articular DDS: Towards Tailored Disease Modifying Therapies

In addition to overcoming the barriers towards intra-articular delivery of DMOADs, there is also a pressing need to reconsider the process for pre-clinical development of intra-articular DDS intended for DMOAD delivery. There has been an increasing evidence suggesting that biological pathways contributing to OA progression are different for each risk factor, and due to such heterogeneities, there exist different patient phenotypes [81]. However, traditional approach to develop DDS for intra-articular DMOAD delivery involves DMOAD selection without considering which patient subgroup will benefit from it and what animal models would be ideal to evaluate a particular DMOAD. We propose a bottom-up workflow for pre-clinical development of intra-articular DDS based DMOADs (Figure 4). Our approach starts with identification of the specific patient phenotype and molecular target, followed by engineering the DDS, and performing safety, pharmacokinetics (PK), biodistribution and efficacy evaluations. We believe that this bottom-up approach would result in tailored DMOAD therapies for specific OA patient subgroups, which can potentially facilitate small sized, and more focused clinical trial design with greater chances of success.

Figure 4. Bottom-up workflow for pre-clinical development of intra-articular DDS based DMOADs.

A bottom-up workflow is proposed to develop tailored DMOAD therapies for specific OA patient subgroups. Our approach consists of following steps: 1) Identify patient sub-group and DMOAD: Imaging techniques can be combined with biochemical and multi-omic techniques to define specific OA subtype with specific structural changes and pathological mechanisms, and to identify the corresponding DMOAD; 2) Select a DDS: DDS can be selected based on two important factors: a) Barriers to overcome to reach the target tissue, and b) Physiochemical properties of the DMOAD; 3) Tune release kinetics and evaluate mechanical stability: DMOAD release could be faster in OA joints compared to healthy joints, and should therefore be evaluated under OA joint conditions . Impact of mechanical loading of joints on the DDS and on the DMOAD release should also be evaluated using a rotational rheometer with strain parameters relevant to human knee joints; 4) Evaluate safety, pharmacokinetics (PK) and local biodistribution: Safety evaluations can be performed using a combination of in vitro, ex vivo and in vivo approaches. PK and local biodistribution studies are critical to understand the joint residence time, maximize localization of the DDS in the target tissue and minimize systemic exposure; 5) Evaluate therapeutic efficacy: Animal model must recapitulate the target OA phenotype and pathophysiology; and 6) Evaluation in large animal models: Safety, PK and efficacy should also be studied in dog, sheep or horse, which mimic human joint anatomy more closely compared to small animals.

Identify Patient Subgroup and DMOAD

Imaging-based scoring systems and quantitative image assessment methods, including radiography, CT, and MRI are already used for OA phenotyping based on structural and compositional changes in joint tissues [82]. Recently, biochemical and multi-omics markers have also been correlated with disease stage or progression [83]. The combination of these two approaches can help define OA subtype with specific structural changes and pathological mechanisms and identify corresponding DMOAD candidates. OA patients with structural progression can be classified into four subgroups including cartilage damage, bone marrow lesions, meniscal tears and extrusion, synovitis, and effusion based on the initial radiograph [82]. MRI can be used to further subtype the cartilage damage group based on measurement of cartilage volume, thickness and denuded surface area [82]. Then each patient subtype can go through biochemical marker analysis such as urinary CTX-II, and serum and urine C1, 2C to identify patients that share similar underlying pathological pathways [83]. Together, all these evaluations can provide great insights into the selection of disease targets and the corresponding DMOAD candidates. In parallel, collaborative efforts from clinicians, molecular biologists, geneticists, and pathologists should be focused towards expanding the list of promising biological targets and DMOADs (See Outstanding Questions). Overall, we suggest narrowing the patient population down to a specific OA phenotype to pinpoint specific molecular targets, and then develop a customized DDS strategy.

Select an Appropriate DDS

Selection of the DDS must consider two important factors: 1) barriers that need to be overcome to reach the molecular target, and 2) physiochemical properties of the DMOAD. Each DMOAD must be delivered to the right molecular target and therefore needs to overcome specific barriers. For example, MK2, a p38 pathway inhibitor, is intended to target synovium to inhibit inflammation [84], which means that particles with a size range of 0.3–10 μm would be an appropriate choice as a DDS [38]. In addition to barriers, physiochemical properties of the DMOAD, including its hydrophobicity and size also define the DDS choice. There are abundant lipids and polymers that can stably encapsulate hydrophobic DMOADs. However, achieving high drug loading and sustained release of hydrophilic drugs can be a huge challenge. For biologic DMOADS such as RNAs or proteins with high molecular weight and charged functional groups, maintaining stability and minimizing systemic immunogenicity can be additional challenges [85]. Therefore, DDS for hydrophilic or biologic DMOADs typically involve more complexity, including chemical modification of the DMOAD. For example, Geiger et al reported intra-articular delivery of IGF-1 by conjugating it to a PEGylated PAMAM dendrimer [70]. However, complicating DDS design can result in scale up and manufacturing issues. Careful consideration therefore must be given to minimize DDS complexity.

Tune Release Kinetics and Evaluate Mechanical Stability

Release kinetics of DMOADs is a critical determinant of the injection frequency required to achieve sustained therapeutic effect. Most of the previously developed DDS for DMOAD delivery have mainly relied on Fick’s diffusion and material degradation to release DMOADs. There have also been some efforts to design intra-articular DDS that exhibit DMOAD release controlled by joint associated reactive oxygen species (ROS), and enzymes such as MMPs [86]. One critical area that has not been explored at all is how local release kinetics of DMOADs in joint impact therapeutic efficacy. We suggest that efforts should be focused on understanding this relationship, as it could maximize the therapeutic efficacy of DMOAD (See Outstanding Questions).

In addition to tuning release kinetics of DMOADs, it is also critical to evaluate the impact of repeated mechanical loading of joints on the DDS and on the DMOAD release kinetics. This is especially critical for DMOADs that are intended for early phase of OA, when patients are physically active. Although mechanical loading has been given due consideration while developing scaffolds for cartilage tissue engineering [87–89], none of the previous work related to DMOAD delivery has looked into the impact of dynamic mechanical loading on structural integrity of DDS or on DMOAD release. Since small animal models cannot recapitulate the strain and stress experienced by human knee joints under mechanical loading conditions, in vitro assays performed using a rotational rheometer or other mechanical testers could be used. These assays can be performed with strain and frequency parameters described previously for resting and running human knee joints [90]. Finally, it is also critical to consider that the requirement of mechanical strength of DDS might be different for early, middle versus late-stage OA due to varying physical activity of patients in different stages.

Evaluate Safety, Pharmacokinetics and Local Biodistribution

If the DDS material has not been previously explored for joint applications, monolayer cultures of human chondrocytes, synoviocytes, osteoblasts/osteoclasts and adipocytes can be used for initial evaluation of biocompatibility. In vitro inflammatory response and oxidative stress of materials may also be evaluated by exposing the synovial macrophages with the test material. ISO 10993–1 provides guidance on possible assays that can be used to demonstrate safety of a medical device or biomaterial [91]. Other than monolayer cell culture, in vitro 3D models can also be used to evaluate DDS in the presence of biochemical cues [92]. Tissue explants can provide valuable information on the whole tissue response to the DDS [93]. Although in vitro and ex vivo models are helpful for initial safety evaluations, in vivo evaluation remains the gold standard and is usually performed in rodents or rabbits. Most published research has considered lethality, body weight, and inflammatory response as the standards for safety evaluation. However, from a clinical translation point of view, safety evaluation should also include sensitization, irritation, material mediated pyrogenicity, acute systemic toxicity, and genotoxicity [91].

Evaluating PK, and local biodistribution of DMOAD loaded DDS is important to determine joint residence time, injection frequency, and to minimize systemic exposure. We suggest that PK and local biodistribution must be evaluated in both healthy and OA animals, as disease pathophysiology can impact the outcome (refer to the barrier section). For PK studies, it is critical to quantify drug levels in the injected joint as well as in plasma/serum. Local biodistribution of the DDS, i.e. localization to specific tissues or cells within the injected joint is critical too; however, most of the previous reports have not focused on this. Finally, we emphasize that evaluating safety, PK, and local biodistribution should be an iterative process. If the formulation results in undesirable or sub-optimal results, researchers should identify the issues and re-engineer the DDS.

Evaluate Therapeutic Efficacy

Small animal models of OA include spontaneous models, transgenic models, and surgical/chemical models [94]. Choosing the right OA model for efficacy evaluation is as important as choosing the right DMOAD or the right DDS. An ideal animal model should recapitulate the specific OA phenotype and pathophysiology that the chosen DMOAD targets. Failure to use the right animal model may lead to false negative/positive results, thereby impeding clinical translation of the system. Since OA is a multi-faceted and heterogeneous disease, none of the animal models can fully recapitulate every aspect of OA [95]. However, each model has its own advantages and disadvantages, which must be weighed carefully against the research need. For example, if the chosen DMOAD is intended to target synovium inflammation, it could be evaluated in collagen-induced arthritis model, that recapitulates synovium inflammation.

The animal model should also ideally simulate the influence of external factors on the joint as well as on the DDS. For example, as obese population climbs up every year worldwide [96], research is being directed towards understanding obesity-induced OA [97]. This has resulted in the development of obesity induced OA models in mice and rats, wherein obesity is induced by using high-fat diet [98]. Such models can result in a robust study designs. One more noteworthy point is to decide if animals should be inactive or physically active during the efficacy evaluation. For example, active animals that are spontaneous wheel-runners could be used for efficacy studies of DMOADs that target early OA. On the other hand, inactive animals can be used to evaluate efficacy of DMOADs that target late-stage OA. However, since small animal models cannot imitate mechanical loading that is experienced by human joints, large animal studies are necessary and should be the logical next step.

Evaluation in Large Animal Models

Rodent and rabbit joints are small compared to humans, and impose strict limits on the total injectable volume of the DDS [99]. Therefore, PK, biodistribution and efficacy studies in small animals cannot accurately predict the outcomes in humans and should only be performed to compare different formulations. Formulations that show promising results in small animal models should be next evaluated in large animal models (ex. dogs, horses, or sheep). At minimal, large animal studies must include both safety and PK evaluations. Also, both single and repeat dose studies are required. Although multiple large animal models of OA are available [100], choosing a model that doesn’t recapitulate the target OA pathophysiology might result in false negative or positive results. Once large animal data is gathered, an investigational new drug (IND) application can be filed with the FDA to advance the formulation into clinical trials.

Other Considerations for Clinical Translation

A common feature of previous clinically tested formulations for OA and other applications is the simple design of their DDS [27,28]. Characterization and validation of complex systems for scale up and manufacturing purposes can be challenging. For example, DDS that requires complex chemistry for synthesis may hit roadblock in terms of achieving reproducibility or minimizing batch-to-batch variations. Innovative DDS designs ar certianly important; however, we must note that complexities will add additional steps and time towards clinical translation. Therefore, innovation must be balanced with simplicity.

Major considerations should also be given to the clinical trial design. Previous clinical trials related to DMOADs have not considered disease heterogeneity. Recently, Pitzalis et al introduced the concept of ‘patient-centric, molecular pathology-driven clinical trial approaches’ in rheumatology [101]. We believe that these principles can also be extrapolated to design clinical trials for DMOADs. Patients recruited for DMOAD clinical trials are often stratified based on the structural changes in joints, as identified by radiographs and are fit into the Kellgren-Lawrence (KL) scale for different severity level [102]. However, minute structural changes may easily evade radiograph diagnosis leading to false negative [101]. Therefore, advanced imaging techniques including MRI and CT scan must be used in addition to the radiograph classification. Patient stratification should also involve OA biomarker analysis such as cartilage oligomeric matrix protein serum level [103]. Although biomarker analysis for OA is still at its infancy stage [83], emerging phenotyping techniques such as genomic, transcriptomic, proteomic and metabolomic technologies can be used for rapid advancement [104]. Another issue with clinical trial design is that the most commonly used primary endpoint is pain [3], which heavily relies on patient reporting and is therefore susceptible to a high degree of variability [3]. For clinical trials that target structural changes in OA, joint space loss, joint space width or cartilage thickness are the primary outcome measures [3]. However, there are also multiple concerns related to each of these metrics [3]. Overall, we emphasize that future clinical trial design should deeply think about endpoints.

Concluding Remarks and Final Perspective

Rapid advancements in our understanding of OA pathophysiology has resulted in multiple promising DMOADs, which have demonstrated encouraging results in clinical studies. Lorecivivint, a novel Wnt-inhibitor, is currently in phase III trial (NCT03928184) and has shown improvements in joint space width and pain relief upon intra-articular injection in OA patients [1]. MIV-711, an oral cathepsin-K inhibitor, also showed significant reduction in bone and cartilage volume loss in a phase II trial (NCT02705625). Lastly, phase II trial of oral Sprifermin (recombinant human FGF18) showed reduction of loss of total femorotibial cartilage thickness (NCT01919164). In conjunction to the blooming investigations of DMOADs, novel intra-articular delivery strategies are also being developed to maximize the efficacy and safety of DMOADs. Numerous efforts have been made to overcome barriers towards intra-articular DMOAD delivery. However, there are still pending questions that need to be answered (see Outstanding Questions). Other than research innovations required to overcome barriers, there is also an urgent need to rethink the pre-clinical development process for DDS based DMOAD therapies. We postulate that patient stratification based on OA phenotype and molecular target is the key to success and should be the first step in the development process (see Outstanding Questions). It will facilitate the development of precisely tailored disease modifying therapies that can be evaluated in small sized, and more focused clinical trials with greater chances of success. Pre-clinical development process should also carefully consider the choice of animal model. Another interesting avenue would be the exploration of DMOAD combinations, especially considering the multifactorial nature of OA. More efforts on the discovery and formulation of gene therapeutics (e.g., siRNA, mRNA) could also lead to meaningful outcomes in OA patients. Considerations must also be given to improve clinical trial design to make them more patient-centric and molecular pathology-driven. We believe that such advances will break open the field of intra-articular DMOAD therapy, enabling clinical translation of multiple diseases modifying therapies that can impact the lives of millions of OA patients around the world.

Highlights.

• Multiple barriers impact intra-articular delivery of DMOADs. Research to overcome these barriers has been growing exponentially, with the development of innovative drug delivery technologies, such as particles that can target specific tissues within joints and can also penetrate through dense cartilage matrix.

• There are still several questions that must be addressed to fully overcome these barriers, which would entail gaining a deeper understanding of interaction of DDS with joint microenvironment.

• With advancements in imaging and omics technologies, OA has begun to be recognized as a heterogeneous disease, with patient subgroups.

• Pre-clinical development process of intra-articular DMOADs should begin with patient stratification and identification of the molecular target as the first step, followed by DDS design and evaluation.

• Critical consideration must also be given to scale up and manufacturing, and clinical trial design.

Glossary

Cathepsin K

a potent cysteine protease capable of degrading collagen and other ECM proteins, thus contributing to cartilage degeneration

Collagen-II

fibrous proteins that form the primary structural element of the cartilage

Disease-modifying osteoarthritis drug (DMOAD)

a class of drugs that goes beyond pain relief or anti-inflammation and targets disease-relevant molecular pathways in OA to attenuate its disease progression

Extracellular matrix (ECM)

a 3-dimensional network providing biomechanical and biochemical support to resident cells; commonly made of collagen, proteoglycan, fibrin, laminin, etc

Hyaluronic acid (HA)

a native component in cartilage made of non-sulfated glycosaminoglycans and is slowly degraded in osteoarthritic patients

Intra-articular injection

a type of local administration method where therapeutics is directly injected in between joints

Kellgren-Lawrence (KL) scale

the most common radiographic classification of OA severity based on features such as osteophyte formation, joint space narrowing, altered bone shape, etc.

Ligament

elastic fibrous tissue that connects bones to bones; the anterior cruciate ligament (ACL) connects articular bone, and its injury is correlated with OA

Liposome

a common artificial drug delivery vehicle consisting of at least one lipid bilayer, created from cholesterol and phospholipid (ex. phosphatidylcholine). Due to its amphiphilic nature, liposomes can encapsulate both hydrophobic and hydrophilic drugs

Matrix metalloproteinase (MMP)

a family of enzymes that degrade most of the ECM components. MMP1 and MMP13 play a major role in osteoarthritis

Meniscus

a crescent-shaped paired fibrocartilaginous connective tissue providing cushioning and stabilization for the cartilage

Opsonin

extracellular proteins (ex. antibody), which attach to the pathogens and foreign particles and trigger phagocytosis by macrophages

Poly(lactic-co-glycolic) acid (PLGA)

a copolymer of lactic and glycolic acids commonly used as a DDS for its excellent biocompatibility and biodegradability

Phage-display peptide

artificially programmable affinity peptides that can bind to certain molecular targets; produced by inserting its sequence of bacteriophages’ genome

Proteoglycan

a major component of ECM, with negatively charged chondroitin sulfates covalently linked to core proteins attached to a hyaluronic acid backbone

Poly(ethylene glycol) (PEG)

a polymer commonly used in drug delivery for its stealth properties and high biocompatibility

Residence time

the timespan for a drug to localize within its target tissue site

Small interfering RNA (siRNA)

~20–24 bp; a short double-stranded non-coding RNA that can induce gene silencing to specific gene targets through the RNA interference pathway

Synovium

aka. the synovial membrane, is a connective tissue lining the inner side of the articular joint capsule. Synovitis, inflammation of the synovium, is a common hallmark of inflammatory osteoarthritis

Viscosupplement

a thick and viscous fluid made of hyaluronic acid that is intra-articularly injected and acts as a lubricant within the knee

WOMAC

a self-administered questionnaire for OA patients to rate 3 categories (pain, stiffness, physical function) on the scale of 0 – 4 from normal to extremely severe