PERSPECTIVES

and/or synovium. These techniques might OPINION be useful for delivering drugs for relieving pain and joint inflammation. However, Cartilage-targeting drug delivery: these approaches do not guarantee drug penetration into cartilage (or other target tissues) or reversible binding of drugs inside can electrostatic interactions help? cartilage. Both mechanisms are necessary to elicit the prolonged biological response Ambika G. Bajpayee and Alan J. Grodzinsky needed for cartilage protection. Abstract | Current intra-articular drug delivery methods do not guarantee Drug penetration and retention inside cartilage is a challenging problem. The sufficient drug penetration into cartilage tissue to reach cell and matrix targets at tissue’s ECM contains densely packed, highly the concentrations necessary to elicit the desired biological response. Here, we negatively charged aggrecan proteoglycans provide our perspective on the utilization of charge–charge (electrostatic) enmeshed within a complex collagen interactions to enhance drug penetration and transport into cartilage, and to network23; the ECM prevents sufficient enable sustained binding of drugs within the tissue’s highly negatively charged drug penetration, thereby enabling rapid clearance of the drug from the joint space24. extracellular matrix. By coupling drugs to positively charged nanocarriers that In this Perpsectives article, we first have optimal size and charge, cartilage can be converted from a drug barrier into describe candidate drugs for the treatment a drug reservoir for sustained intra-tissue delivery. Alternatively, a wide variety of of OA and PTOA, and then focus on drugs themselves can be made cartilage-penetrating by functionalizing them with mechanisms by which charge–charge specialized positively charged protein domains. Finally, we emphasize that interactions can increase drug penetration, appropriate animal models, with cartilage thickness similar to that of humans, must transport kinetics and retention within charged, avascular tissues such as be used for the study of drug transport and retention in cartilage. cartilage. We compare three approaches to intra-articular cartilage-targeted No disease-modifying osteoarthritis drugs in OA‑associated cartilage pathogenesis5. delivery, and end with a discussion on the (DMOADS) are currently available. Several Drug penetration into cartilage is especially appropriate animal models to use for testing drugs have potential to inhibit cartilage important following traumatic joint injury, these systems. degeneration associated with osteoarthritis which can result in damage to articular (OA) and post-traumatic osteoarthritis cartilage, subchondral bone and nearby Candidate disease-modifying drugs (PTOA), and to promote cartilage repair1; soft tissues, initiating a sequence of Current therapies for OA provide only however, none of these drugs have yet inflammatory events that can progress to short-term relief of pain and inflammation translated to clinical practice, owing in part PTOA6. Biopsy-obtained samples of cartilage (for example, analgesics and hyaluronic to the lack of effective delivery systems that from anterior cruciate ligament injury have acid lubricants), but no protection against enable local, safe administration in low revealed degradative changes to cartilage further degeneration of cartilage and OA doses without off-target effects2,3. Direct as early as 3 weeks after injury, including progression25. Several therapeutics have intra-articular administration of drugs can loss of superficial zone proteoglycans and been identified as having the potential for minimize adverse systemic side-effects4. cell viability, even in cases where there is no disease-modifying inhibition of cartilage But even intra-articular injection remains obvious damage to cartilage or its collagen breakdown, including anticatabolic inadequate, as small compounds and large network, as visualized by arthroscopy7. glucocorticoids (such as dexamethasone and macromolecules are rapidly cleared from New drug-delivery systems have been triamcinolone)12,26,27, cytokine blockers28,29, the joint space via subsynovial capillaries proposed for sustained delivery in the proanabolic growth factors (including and lymphatics, respectively. For example, synovium and synovial fluid using polymeric insulin-like growth factor (IGF) 1 (REFS 30,31), the mean half-lives of NSAIDs in the nanoparticles8–11, microparticles12,13, fibroblast growth factor (FGF) 18 (REFS 32,33) synovial fluid are only 1–4 h (REF. 4). As liposomes14,15, drug-loaded hydrogels16–19, and bone morphogenetic protein (BMP) 7 a result, multiple injections of high-dose phase transitioning elastin-like (REF. 34)) and chondrogenic biomolecules35. drugs are sometimes used in attempts to polypeptides20, silk constructs21, and Given that OA affects the entire joint, suppress pain, inflammation and cartilage electrospun fibres22. These drug carriers have DMOAD development and associated destruction, an approach that can cause prolonged residence times due to their large clinical trials have targeted cartilage systemic toxicity3. size (micron) or viscous and/or aggregating breakdown (with protease or cytokine Drugs need to penetrate the full depth properties that prevent them from leaving blockers), bone remodelling (with of cartilage to reach the chondrocytes and the joint space rapidly, thereby enabling bisphosphonates, BMP7 or calcitonin), extracellular matrix (ECM) targets involved rapid drug release within the synovial fluid and synovial and inflammatory mediators

NATURE REVIEWS | RHEUMATOLOGY VOLUME 13 | MARCH 2017 | 183 ©2017 Mac millan Publishers Li mited, part of Spri nger Nature. All ri ghts reserved.

PERSPECTIVES

Table 1 | Examples of potential drugs for OA treatment under experimental or clinical trial testing Drug type Drug action Examples Molecular Mechanism of therapy Target location and/or target weight inside joint NSAIDs Pain relief • Ibuprofen <500 Da Inhibit COX enzymes • Vasculature of the joint • Naproxen capsule and cartilage– • Celecoxib bone interface • Free nerve endings Monoclonal antibodies Biologic agents for • Tanezumab ~150 kDa Bind to and inhibit NGF, of sensory neurons in against NGF pain relief • Fluranumab which is produced by soft tissues (e.g. patella OA synovial cells and ligament and below the chondrocytes and acts synovial layer) directly on sensory neurons Monoclonal antibodies Biologic agents as TNF inhibitors* ~150 kDa Directly bind target Cytokine targets against iflammatory DMOADs • Infliximab (except cytokines, preventing them hypothesized to be cytokines • Adalimumab Etanercept, from binding with their in the synovium, the • Etanercept ~50 kDa) respective cell-surface synovial fluid and found receptors to initiate throughout the full IL-1β inhibitors* signalling depth of the cartilage • Canakinumab extracellular matrix Receptor antagonists Biologic agents as IL‑1 receptor ~17 kDa Competitively bind with DMOADs antagonists* cell-surface cytokine • Anakinra receptors thereby blocking cytokine activity Glucorticoids Pain relief at Salts of dexamethasone, <1 kDa Bind with intracellular Full depth of cartilage as high doses (and triamcinolone and glucocorticoid well as neighbouring soft anticatabolic prednisone receptors and inhibit tissues and synovium effects in cartilage cytokine-induced catabolic at low doses) activity Growth factors Biologic agents as IGF‑1, FGFs, BMPs 10–20 kDa Bind with cell surface Full depth of cartilage, DMOADs growth factor receptors to meniscus and other stimulate repair tissues Protease inhibitors and DMOADs Inhibitors of MMPs, <1 kDa Bind with the catalytic zinc Full depth of cartilage, pro-protein convertase aggrecanases atom at the MMP active synovium and joint blockers (ADAMTS‑4, site (for MMP inhibitors) capsule space ADAMTS-5), cathepsins, to inhibit cartilage ECM PACE4, and others breakdown Viscosupplements Pain relief and joint Hyaluronan‡. lubricin 2–6 MDa Intended to restore Synovial fluid, joint lubrication (proteoglycan 4)and joint lubrication, and capsule, synovial others hypothesized to bind with membrane and superficial CD44 receptors to induce zone of cartilage chondroprotection ADAMTS, a disintegrin and metalloproteinase with thrombospondin motifs; BMP, bone morphogenetic protein; COX, cyclooxygenase; DMOAD, disease-modifying osteoarthritis drug; ECM, extracellular matrix; FGF, fibroblast growth factor; IGF‑1, insulin like growth factor 1; MMP, matrix metalloproteinase; NGF, β-nerve growth factor; OA, osteoarthritis. *Currently used for systemic treatment of rheumatoid arthritis. ‡Approved but no longer recommended for patients with symptomatic knee OA, according to American Academy of Orthopaedic Surgeons 2013 evidence-based guidelines82.

(with cytokine blockers)36. TABLE 1 is a For example, a multicentre, randomized, Another randomized, double-blind, representative list of such therapeutics that are double-blind, placebo-controlled study placebo-­controlled, multiple-dose study currently being considered for OA treatment. (NCT00110916 (REF. 38)) was performed (NCT00110942) used subcutaneous injection Biologic agents such as monoclonal to evaluate the clinical response, safety or infusion of a monoclonal antibody antibodies against IL‑1β (canakinumab) and and tolerability of a single intra-articular (AMG 108) that binds the IL‑1 receptor TNF (infliximab, adalimumab), and other injection of anakinra (an IL‑1 receptor type 1 (IL‑1R1), thereby inhibiting the anti‑IL‑1 or anti-TNF agents (anakinra, antagonist (IL‑1Ra), molecular weight activity of IL‑139. The results showed etancercept), have been used successfully ~17 kDa) in patients with symptomatic knee statistically insignificant but numerically for the treatment of rheumatic diseases via OA. Although significant improvement was greater improvement in WOMAC (Western systemic delivery. Notably, monoclonal observed at day 4, anakinra did not improve Ontario and McMaster Universities Arthritis antibodies and similarly sized therapeutics are OA symptoms after 1 month when compared Index) pain score compared with the placebo probably much too large to penetrate cartilage with placebo38. The results suggest that the group, but the clinical relevance was stated to sufficiently before being rapidly cleared from drug had cleared out from the joint space be unclear. The authors stated that it was not the joint24,37. Although some of these agents rapidly following intra-articular injection, possible to evaluate the penetration of AMG are being considered for intra-articular and furthermore showed a serum half-life 108 into the deeper cartilage layers, and that delivery to treat OA, clinical trials have of only 4 h. The investigators speculated the availability of the drug to chondrocytes in lacked evidence of either sustained benefit or that multiple injections would be needed in cartilage remains a possible limitation of this effective cartilage targeting3. any attempt to achieve cartilage protection. strategy for IL‑1 inhibition.

184 | MARCH 2017 | VOLUME 13 www.nature.com/nrrheum ©2017 Mac millan Publishers Li mited, part of Spri nger Nature. All ri ghts reserved. ©2017 Mac millan Publishers Li mited, part of Spri nger Nature. All ri ghts reserved.

PERSPECTIVES

Collagen network filled with To summarize thus far, certain drugs 300MDa aggrecan aggregates intended for relief of pain and general joint inflammation might do well with SZ Chondrocyte delivery and retention in the synovium or synovial fluid. However, to achieve cartilage protection (that is, to protect MZ chondrocyte viability, inhibit cartilage matrix degradation and stimulate cartilage matrix biosynthesis), preclinical and clinical studies to date strongly suggest that appropriate therapeutics must be delivered DZ to chondrocytes (especially in the middle and deep zones of cartilage5) or to cartilage Increasing concentration of aggrecans concentration Increasing matrix-associated targets.

CZ Delivering drugs to cartilage Cartilage: a barrier to drugs Blood vessels Articular cartilage is a highly complex,

zone avascular, alymphatic and aneural tissue Collagen whose matrix is made of a dense network Subchondral of collagen fibrils (50–60% dry weight of tissue), aggrecan proteoglycans that Hyaluronan contain highly negatively charged glyco­s­aminoglycan (GAG) chains (30–35% tissue dry weight) and dozens of additional extracellular macromolecules, which are continuously synthesized by a low density of chondrocytes (1–5% tissue dry weight)42. The collagen fibril network Aggrecan monomer (mostly type II collagen with some type IX and XI collagen43) has an approximate Chondroitin sulfate GAG chain ~25kDa pore size of 60–200 nm (REF. 44). Collagen fibrils are aligned parallel to the surface Core protein G1 in the superficial zone (about 10–20% of total cartilage thickness), but are randomly Link protein oriented in the middle zone (40–60% Keratan sulfate tissue thickness) and perpendicular GAG chains Aggrecan 2–4 nm to the subchondral bone in the deep aggregate zone (30–40% tissue thickness). The ~300 MDa ~200–400 nm collagen network is filled with ~300 MDa ~3 MDa aggregates formed mainly of aggrecan; each aggregate comprises a central hyaluronan Figure 1 | Dense meshwork of type II collagen and aggrecan makes cartilage a barrier to drug GAG chain to which as many as one penetration. The density of aggrecans increases with depth into cartilageNature Reviews towards | Rheumatologythe deep zone hundred 2–3MDa aggrecan monomers (DZ). The superficial zone (SZ) forms 10–20% of total cartilage thickness, the middle zone (MZ) 40–60% are noncovalently bound via G1 binding and the DZ 30–40%. The calcified zone (CZ) and subchondrial zone are also depicted. Aggrecan aggre- domains, an interaction further stabilized gates are ~300 MDa macromolecules comprising hyaluronan (a long, central glycosaminoglycan by a link protein (FIG. 1). The sulfated GAG (GAG) chain) and 100 or more ~3 MDa aggrecan monomers, which are bound non-covalently to hya- chains covalently linked to the aggrecan luronan via their G1 globular domains and further stabilized by a link protein. Each aggrecan monomer monomers are spaced only 2–4 nm apart has negatively charged chondroitin and keratan sulfate GAG chains that are separated from each along the monomer core protein45. Thus, other by 2–4 nm along the aggrecan core protein. Parts of this figure are reproduced with permission from Nia, H. T. et al. High-bandwidth AFM-based rheology reveals that cartilage is most sensitive to these bottle-brush structured aggrecan monomers are so densely packed within high loading rates at early stages of impairment. Biophys. J. 104, 1529–1537 (2013) (REF 81). the collagen network that the GAG chains on adjacent aggrecan monomers are essentially as close to each other as A new class of RNA interference (MMP) and a disintegrin and metallo­ GAG chains along the core protein46. (RNAi)-based therapeutics has also proteinase with thrombospondin motifs Taken together, this matrix composition emerged to target transcription factors (ADAMTS) aggrecanases, but these presents substantial steric hindrance to the (for example, NF‑κB and HIF‑2α) and therapeutics can become clinically penetration of therapeutic molecules. In their target genes, including those relevant only if there is a way to deliver addition, the density of aggrecan increases encoding matrix metalloproteinase them to chondrocytes40,41. with depth into cartilage, which further

NATURE REVIEWS | RHEUMATOLOGY VOLUME 13 | MARCH 2017 | 185 ©2017 Mac millan Publishers Li mited, part of Spri nger Nature. All ri ghts reserved. ©2017 Mac millan Publishers Li mited, part of Spri nger Nature. All ri ghts reserved.

PERSPECTIVES

reduces the effective pore size and restricts Drug transport into cartilage intra-cartilage therapeutic levels before the the ability of solutes to penetrate and diffuse Drug penetration and retention in cartilage drugs are cleared from the joint space47. within the tissue. Given that the majority depends on two competing rates of Clearance is rate-limited first by elimination of chondrocytes reside in the middle and transport (FIG. 2): first, the net flux of drug through the synovial membrane and then deep zones of the tissue, drug delivery to carriers entering cartilage from synovial by the systemic circulation48. Detailed chondrocytes is a challenge, and avascular fluid, NEntry; and second, the rate of exit from pharmacokinetic models of the escape cartilage is clearly a barrier to drug and/or the lymphatics and subsynovial capillaries, kinetics of drugs from the synovial cavity 4,48 drug-carrier entry. NExit. NEntry should be fast enough to achieve can be found elsewhere in the literature . a b T1 CSF

Synovial Bone Diffusion into cartilage membrane

* CM CC T2 CSF

CC Intra-articular CM * injection Cartilage T3

CSF CC * CM

NENTRY Drug concentration T4 Diffusion out of cartilage NEXIT

CC Synovial fluid CSF * CM T5

CC * CM CSF Membrane Synovial Cartilage Bone fluid c Diffusion into cartilage Diffusion out of cartilage

Cartilage Synovial fluid

* Drug concentration

τeff

T1 T2 T3 T4 T5 Time Figure 2 | Distribution of drugs or drug carriers inside the joint space fol- drug-carrier concentration profiles at various time points (T1 to T5) is shown lowing intra-articular administration. The concentration of injected drugs during drug accumulation into cartilage andNature depletion Reviews from cartilage | Rheumatology (part b). in the synovial fluid (CSF) is assumed to be homogenous throughout the joint These chosen time points are also illustrated in a graph depicting the drug space shortly after injection (for example, due to joint flexing). Penetration of concentration in the cartilage or synovial fluid over time, following injection the drug into cartilage (NEntry) competes with clearance of the drug through (part c). The time period during which the drug stays above the critical thera- the synovium membrane into the lymphatics and vasculature (NExit) (part a). peutic level inside cartilage (denoted by *) is denoted as τeff, corresponding to A simplified one-dimensional model depicting the transient drug or the period during which the drug is effective in eliciting a biological response.

186 | MARCH 2017 | VOLUME 13 www.nature.com/nrrheum ©2017 Mac millan Publishers Li mited, part of Spri nger Nature. All ri ghts reserved. ©2017 Mac millan Publishers Li mited, part of Spri nger Nature. All ri ghts reserved.

PERSPECTIVES

The manner in which electrically neutral Utilizing electrostatic interactions well as the fixed-charge groups of the ECM drugs and/or drug carriers distribute The high density of negatively charged (for example, the charges on GAGs). Thus, between regions of the joint space following GAGs inside cartilage provides a unique under physiological conditions, Na+ ion intra-articular injection are depicted opportunity to use electrostatic interactions concentration is higher inside cartilage while schematically in FIG. 2. To reduce the to augment transport, uptake and Cl− ion concentration is lower compared mathematical complexity, this scheme binding of drugs and drug-carriers. Such with their concentrations in the synovial has been simplified to a one-dimensional interactions, however, have not yet been fluid at equilibrium (FIG. 3b). transport system and depicts the transient fully exploited for local intra-articular Transport of large-molecular-weight concentration profiles of the drug (or delivery (BOX 1). Three complementary solutes into cartilage is sterically hindered, drug carrier) in these regions at various mechanisms have been identified by typically resulting in partition coefficients time-points during its accumulation (FIG. 2b) which electrostatic interactions can enable <1. For example, Maroudas et al. showed and depletion (FIG. 2c) inside cartilage. cartilage-­targeted drug delivery and that serum albumin (molecular weight Intra-articular injection immediately retention (BOX 2). 69 kDa, diameter ~7 nm, isoelectric point increases the synovial fluid concentration (pI) 4.7) is sterically hindered in normal (REFS 49,51) (CSF) of the drug or drug-carrier. Using Donnan partitioning. Within cartilage, human cartilage, with k <0.05 . the simplest model in which drug or the partition coefficient of a solute, k, is Similarly, neutravidin (66 kDa, electrically drug-carrier concentration becomes defined as the equilibrium concentration neutral) has k = 0.5 in normal cartilage24. By uniformly distributed in the synovial of unbound, free solutes inside cartilage comparison, avidin (66 kDa, net charge +20, fluid volume (for example, by flexing the normalized to the solute concentration pI = 10.5), the same-sized positively charged

joint), the CSF at the interface of synovial in the surrounding bath (that is, synovial counterpart of neutravidin, has a much fluid and cartilage can be assumed to be fluid)49. k depends on solute size, charge and higher partition coefficient (k = 6)24. Most approximately equal to that at the interface the composition of the cartilage ECM. importantly, when a highly positively charged of synovial fluid and synovial membrane. A small electrically neutral solute that is not drug (or a potential drug carrier such as Initially the drug concentration inside sterically hindered by cartilage ECM will avidin) is injected intra-articularly47,52,

cartilage (Cc) will continue to increase with have a partition coefficient of ~1: that is, there is an immediate sharp increase

time even as CSF begins to decrease with when in equilibrium, solute concentration in its concentration just inside the clearance from the synovial fluid. When in the tissue and the surrounding synovial cartilage due to Donnan partitioning at the drug concentration in the synovial fluid is nearly equal. The high negative the synovial fluid–cartilage interface50 fluid finally becomes lower than that inside fixed-charge density of GAGs inside (FIG. 3a). As depicted in FIG. 3a, the resulting cartilage, a net outward diffusion from cartilage results in a drop in the electrical increased concentration (from C to kC) cartilage back into synovial fluid follows potential (ΔΦ) at the tissue interface, causing causes a steep intra-tissue concentration (FIG. 2c), unless there is a mechanism by a strong, inwardly pointing electric field gradient (from the superficial zone inward), which the drug or drug carrier can bind (FIG. 3a) that enhances transport of positively which greatly accelerates transport of to sites inside cartilage. From standard charged species into cartilage and excludes positively charged drugs and/or drug diffusion theory, the diffusion time across penetration of negatively charged solutes. carriers deeper into the negatively charged cartilage is proportional to the square of This intra-tissue distribution of charged cartilage. This accelerated transport enables

the cartilage thickness (Lc); hence, cartilage solutes within charged tissues is quantified drugs and/or drug carriers to penetrate into thickness is extremely important to the by Donnan’s theory42,50,which states that, cartilage faster than their clearance rate success of intra-cartilage delivery (see in equilibrium, all freely moving charged from the synovial fluid. For example, avidin discussion of animal models, below). The solutes will distribute (that is, ‘partition’) into fully penetrates through 1 mm‑thick bovine time period during which the drug stays charged tissues on the basis of the difference cartilage in <24 h, whereas neutravidin only above the critical therapeutic level inside between the mean electrical potential partially penetrates by 4 days; avidin also FIG. 2b–d cartilage (τeff, denoted by * in ) inside the tissue compared with that of the shows 400-fold higher uptake into cartilage corresponds to the period during which surrounding bath and, additionally, that the compared with neutravidin24. the drug is effective in eliciting the desired net charge inside the tissue must be zero, biological response. including the charges of all mobile solutes as Weak reversible binding. Current research in intra-articular delivery has sometimes focused on using strong Box 1 | Why intra-cartilage delivery? binding mechanisms (for example, covalent bonds8,53) to increase the residence time • Potential disease-modifying osteoarthritis drugs (DMOADs) are in preclinical development, of drug carriers inside cartilage. However, but cartilage-targeted delivery methods for intra-articular delivery in humans are lacking such strong-binding mechanisms would • Drugs injected directly into the joint have short residence times as they are rapidly cleared via markedly slow the penetration of these lymphatics or sub-synovial capillaries carriers into human (and large animal) • The dense matrix of cartlilage prevents the drug penetration necessary to elicit the desired cartilage as the carriers would be trapped biological response in the surface layers well before reaching • In the absence of cartilage-targeted intra-articular delivery, multiple injections of high-dose middle zone and deep zone targets drugs might cause off-target effects and even systemic toxicity (BOX 3). More precisely, diffusion-reaction • To reach chondrocyte and matrix targets throughout the full thickness of cartilage, drug carriers transport times are inversely proportional are needed that can penetrate to the deep zone and bind to cartilage matrix, and thereby to the effective diffusivity of drug carriers, provide sustained intra-tissue delivery of therapeutics which would be decreased by orders of

NATURE REVIEWS | RHEUMATOLOGY VOLUME 13 | MARCH 2017 | 187 ©2017 Mac millan Publishers Li mited, part of Spri nger Nature. All ri ghts reserved. ©2017 Mac millan Publishers Li mited, part of Spri nger Nature. All ri ghts reserved.

PERSPECTIVES

Box 2 | How can electrostatic interactions help? drug transport, completely independent of electrostatic effects. However, several Cationic carriers rapidly penetrate negatively charged cartilage, resulting in high uptake, studies have reported that the effects of binding to extracellular matrix components and sustained drug delivery to chondrocytes dynamic loading increase transport by only Transport rate and uptake into cartilage about twofold for large macromolecules • Electrostatic interactions cause a sharp increase in concentration of cationic carriers at the compared with that in the non-loaded synovial fluid–cartilage interface (called Donnan partitioning) following intra-articular injection condition56,57. This approximately twofold • This Donnan partitioning causes steep intra-cartilage concentration gradients that accelerate increase was reported when cartilage plugs transport and enable high uptake of cationic carriers into cartilage before they exit the were subjected to continuous cyclic loading synovial fluid for several hours, a procedure not practical Depth of penetration and binding in clinical situations. Hence, dynamic • Electrostatic binding is weak and reversible (that is, carriers rapidly unbind after initial binding loading has a smaller effect on transport with negatively charged groups); thus, cationic carriers continue to diffuse throughout the full of small molecules compared with passive thickness of cartilage diffusion. By contrast, the electrostatic • Despite weak binding, the high negative fixed charge density of aggrecan glycosaminoglycans effects described here can enable increases in inside cartilage greatly increases the residence time of cationic carriers intra-cartilage concentration of 10–100‑fold, • In the early stages of OA, despite the loss of some GAGs, the remaining negative charges inside as has been reported for various cationic cartilage still provide sufficient binding sites for cationic carriers solutes in cartilage56–58.

Drug delivery approaches magnitude due to tight binding. By contrast, within cartilage) and that of the drug and/or Cartilage-targeting drug carriers nonspecific electrostatic interactions drug carrier molecule. For example, the The current clinical standard for between positively charged drugs and/or heparin-binding domain of heparin-binding intra-articular delivery is direct injection drug carriers and negatively charged (HB)‑IGF‑1 binds with a higher affinity into the joint. New delivery approaches cartilage ECM lead to weak and reversible to heparan sulfate GAGs (KD= 21 nM) being investigated in clinical trials utilize binding, which provides the distinct than to chondroitin sulfate GAGs micron-sized, impenetrable, non-binding 31 advantage of enabling drug carriers to (KD = 160 nM) . Similarly, avidin binds particles that remain suspended in rapidly penetrate through the full thickness weakly to chondroitin sulfate GAGs. the synovial fluid for sustained drug of cartilage. Such cationic particles will However, this weaker binding is release into the joint space. However, a be attracted to and could weakly bind compensated by the much higher density considerable fraction of the released drug negatively charged GAG constituents of chondroitin sulfate GAGs than heparan might be cleared from the joint before inside cartilage. This weak and reversible sulfate GAGs in cartilage (500–1,000 fold)31. entering cartilage. Hence the available ionic binding has a correspondingly high By contrast, although the small cationic drug concentration inside cartilage, Cʹ, will dissociation constant, KD, so the carriers also peptide therapeutic Pf‑pep (Arg-Tyr- be lower than the concentration of drug rapidly unbind from their intra-tissue Lys-Arg-Thr, 760 Da, net charge +3, pI ~11) injected into the joint, C. The time required binding sites. If the particles are small was found to partition upward (k ~3.5) into to reach intra-cartilage therapeutic levels (τ1) enough such that they are not subject to cartilage, this peptide did not bind inside can be very long (if ever attained) (FIG. 4a). steric hindrance by the cartilage ECM, cartilage and rapidly diffused out of the Most currently explored sustained-release they will continue to diffuse through the tissue, thereby preventing its intra-cartilage drug delivery systems fall into this category; cartilage and penetrate deeper into retention at levels needed for sustained examples include triamcinolone-loaded the tissue, given the inward concentration therapeutic effect54. Thus partitioning and PLGA (poly lactic-co‑glycolic acid) gradient induced by Donnan partitioning at binding are two independent mechanisms microspheres12, carriers or drugs crosslinked the superficial zone, as described above. that affect solute uptake and retention to either exogenous hyaluronan particles59 inside cartilage in very different ways or endogenous hyaluronan within the joint9, High intra-tissue binding site density. (FIG. 3c). Separate experiments must be and elastin-like polypeptides for delivery of The high density of GAGs in cartilage performed to test whether electrostatic IL‑1Ra60. These drug delivery systems are provides a high density of binding sites interactions can simultaneously enable both most relevant if target sites are mainly in the for certain positively charged solutes, upward partitioning and binding24. synovial fluid or synovium, such as when which greatly increases their intra-tissue mediating pain and inflammation, but are residence time, despite their weak binding. Effects of dynamic loading not very effective for targeting chondrocytes For example, avidin remains bound within Dynamic loading of cartilage, as would occur unless extremely high drug doses are used. cartilage for several weeks owing to a very during walking, running or jumping, A second approach utilizes drug carriers high intra-cartilage binding site density might also affect drug uptake. These of varying sizes functionalized to bind to (FIG. 4b) (NT ~2,900 µM) even though its binding effects are independent of the electrostatic cartilage surfaces . Such surface 24 affinity is very weak (KD = 150 µM) . interactions discussed above. Joints are binding prevents these carriers from However, although electrostatic mechanically loaded across a wide spectrum penetrating deeper into the tissue, although interactions result in upward partitioning of of frequencies (loading rates) depending drugs released from them could penetrate cationic peptides into cartilage, they do not on the type of physical activity, which effectively depending on their properties. guarantee binding to the matrix. Binding compresses the cartilage55. This dynamic As a result, drug concentrations could reach depends on the precise chemical structure loading of joints results in fluid flow within intra-cartilage therapeutic levels in a shorter of the binding site (for example, the GAGs cartilage that could potentially enhance time, τ2, compared with drug-release from

188 | MARCH 2017 | VOLUME 13 www.nature.com/nrrheum ©2017 Mac millan Publishers Li mited, part of Spri nger Nature. All ri ghts reserved. ©2017 Mac millan Publishers Li mited, part of Spri nger Nature. All ri ghts reserved.

PERSPECTIVES

a Synovial fluid Cartilage Bone b Synovial fluid Cartilage Bone have failed clinical trials65 owing to a lack of tissue-targeted delivery and resulting – – – kAv+ CAv+ CAv+ systemic adverse effects. – – kNa+ CNa+ CNa+ C – – – Cl- Cartilage-targeting drugs ΔΦ k C – – Cl– Cl– Other approaches have been explored for E Impermeable Impermeable targeting chondrocytes inside cartilage, – – – including gene delivery66 (which often involves use of cationic viruses) and the use of ultrasonography to increase drug 67 c transport rates . Attention has also focused Avidin on the addition of cationic domains to Positively charged HB-IGF-1 known protein therapeutics (for example, Weak binding Positively charged K ~150 µM D Strong binding growth factors) to enable electrostatic

KD~160nM penetration, binding and retention inside cartilage in a manner similar to that pictured Pf-pep Positively in FIG. 4c. Clues for this approach were charged garnered from the FGF family, one of the No binding first ligands discovered to have a cationic heparin-binding domain68. This domain binds heparan sulfate GAG chains on the surface of cells in cartilage, increasing the affinity between FGF family ligands | Figure 3 Electrostatic (charge–charge) interactions cause Donnan partitioning but not neces- and their cell-surface receptors, crucial for sarily drug binding to cartilage matrix. a | The high negative fixed-chargeNature Reviews density of| Rheumatology glycosamino- stabilizing this formation and initiating glycans (GAGs) inside cartilage results in a decrease in the electrical potential (ΔΦ) from the synovial 69 fluid into cartilage, owing to a strong, inwardly pointing electric field (E) that enhances transport of cell signalling . positively charged species into cartilage and diminishes penetration of negatively charged solutes In addition to facilitating growth factor such that the total net charge inside the cartilage is zero. b | The concentration of positively charged signalling, heparin and heparan sulfate are + − Na partitions upward (from CNa to kNaCNa), while negatively charged Cl partitions downward (from of particular interest in the realm of growth

CCl to kClCCl). Similarly, positively charged drug carriers (for example, avidin) partition upwards into factor delivery owing to the cationic nature of negatively charged cartilage. k is the partition coefficient and C is the concentration of the solute heparin-binding domains. The existence particle within the synovial fluid. c | The small cationic peptide Pf‑pep does not bind inside cartilage of this domain on FGF‑18 was partly the and hence rapidly diffuses out54. Avidin binds weakly and reversibly with the negatively charged motivation for utilizing this growth factor for 33 GAGs with a dissociation constant (KD) of 150 µM owing to charge–charge interactions (note that a ongoing OA clinical trials high dissociation constant implies weak binding)24. The cationic heparin-binding (HB) domain of Although some proteins (such as the FGF heparin-binding insulin-like growth factor 1 (HB‑IGF‑1) binds comparatively more strongly with chon- family and vascular endothelial growth factor droitin sulfate GAG chains than does avidin, although penetration of HB‑IGF‑1 into cartilage is still (VEGF)) have naturally occurring cationic dramatically high31. domains that bind to heparin and heparan sulfate, such heparin-binding domains can also be attached to other molecules, (FIG. 4c) impenetrable non-binding carriers (τ2 < τ1, concentrations in the shortest time . resulting in fusion proteins such as HB‑IGF‑1 FIG. 4b). This method might be best suited for For example, avidin has been shown to that are now known as heparin-binding delivery of novel proteoglycan 4 (PRG4)-like have optimal size and charge properties drugs. Studies in vitro, as well as in animal proteins for surface lubrication61. for intra-cartilage drug delivery24,47. models30,31, revealed that intra-articular In contrast to impenetrable particles, When conjugated with dexamethasone, injection of HB‑IGF‑1 resulted in longer positively charged nanosized carriers it rapidly penetrated into full-thickness retention and bioactivity in cartilage, as well <10 nm diameter (cationic drug nanocarriers) cartilage explants, releasing the drug as enhanced local delivery to chondrocytes, can penetrate past the superficial zone inside, which significantly suppressed compared with IGF‑1. Furthermore, of the cartilage24 (FIG. 4c). Carriers with IL‑1‑induced GAG loss over 3 weeks HB‑IGF‑1 also binds to, and is primarily similar diameter but longer length (for compared with free drug63. In another retained by, chondroitin sulfate35, a promising example, rod or chain structures) might study, cationic moieties were incorporated finding given the much higher concentration also penetrate cartilage tissue owing to into DOTAM (1,4,7,10‑tetraazacy- of chondroitin sulfate GAGs in cartilage matrix tortuosity62. The positive charge clododecane‑1,4,7,10‑tetraacetic acid compared with heparan sulfate GAGs. Thus, of nanocarriers drastically increases amide)-based nanocarriers functionalized cationic drugs take advantage of the same their partitioning and results in steep with the cathepsin D inhibitor pepstatin electrostatic interactions described above for intra-tissue concentration gradients and A, and demonstrated retention in mouse cationic drug carriers, interactions that enable accelerated transport. Mediated by weak knee joints64. Thus, this approach could upward partitioning, accelerated transport reversible ionic binding, this approach enable intra-cartilage delivery of potential and retention of bioactive growth factors can yield full depth penetration and DMOADs, which has remained a challenge. inside cartilage. Furthermore, an entire class retention of drug carriers, resulting in Such an approach might also offer a unique of glycoadherins and chondradherins could therapeutic levels of intra-cartilage drug opportunity to re‑examine OA drugs that be used to functionalize drugs. For example,

NATURE REVIEWS | RHEUMATOLOGY VOLUME 13 | MARCH 2017 | 189 ©2017 Mac millan Publishers Li mited, part of Spri nger Nature. All ri ghts reserved. ©2017 Mac millan Publishers Li mited, part of Spri nger Nature. All ri ghts reserved.

PERSPECTIVES

Box 3 | Strong binding hinders diffusive transport Late-stage OA is characterized by extensive damage to the collagen network The ability of binding interactions to promote transport of carriers through tissues is paradoxical of cartilage and other soft tissues, as well because binding enhances retention but hinders diffusive transport. Using strong binding as substantial loss of GAGs from cartilage. mechanisms such as covalent bonds for increasing drug retention inside cartilage can sound These changes are accompanied by attractive; however, very strong or irreversible binding can dramatically slow down transport of carriers as they get bound/trapped in the surface layers of cartilage, preventing them from episodic synovitis, osteophyte formation 25 penetrating further to reach their target sites. and subchondral sclerosis . Although this In constrast, nonspecific electrostatic interactions between positively charged carriers and more extreme degradation of cartilage the negatively charged ECM leads to weak and reversible binding, which provides the makes the tissue much more permeable to distinctive advantage of allowing drug carriers to rapidly penetrate through the full thickness larger sized drugs (perhaps even ~150 kDa of cartilage. antibodies), this advanced stage of disease might be irreversible, and treatments are probably limited to symptomatic relief of the C-terminal peptide of chondroadherin Experiments utilizing avidin showed that pain and inflammation36. However, GAG selectively binds to heparan sulfate chains69; its concentration inside partially degraded chains are present in the menisci, ligaments similarly, the GAG-binding domain of cartilage (40% depletion of GAG chains) and in lower concentrations in tendons. prolargin (also termed PRELP) can be was 25 times higher than that in the bathing Additionally, lubricin glycoproteins are found fused with drugs to enable binding within medium24. Thus, utilizing charge–charge in the superficial zone of cartilage and in cartilage ECM70. interactions for targeting and retaining the surface layers of the synovial membrane, positively charged drugs and/or drug carriers fat pads and other gliding joint tissues, Delivery to damaged cartilage is feasible even with partial GAG loss. At providing additional natural reservoirs for In the early stages of cartilage degradation, this stage of disease, before overt collagen positively charged drug carriers. Hence, a window of opportunity exists for drug fibrillation, pharmacological intervention attaching a cationic domain to pain and delivery6 when there might be some but not could delay, prevent or even reverse inflammation relievers might still enhance yet complete loss of GAGs from cartilage7. progression of OA or PTOA 71,72. their residence time in the joint through binding to negatively charged molecules in the synovial fluid, fat pads and synovium. Glossary Appropriate animal models Therapeutic levels inwardly pointing electric field that enhances transport of Any perspective on developing new The drug doses necessary to elicit the desired biological positively charged species into cartilage and excludes drug-delivery systems must include a response. For a particular drug this level can be estimated penetration of negatively charged species such that the net using a combination of in vitro assays and in vivo charge inside the cartilage is zero. Thus, the concentration discussion of animal models used to pharmacokinetic and pharmacodynamics studies. of positively charged drug can increase dramatically investigate in vivo biological responses (i.e. partition upwards) across the interface as the drug and transport kinetics. Multiple studies of Diffusion time enters the negatively charged cartilage. intra-articular delivery approaches have Time for diffusion (τ) of a drug into cartilage of thickness 73,74 2 utilized mouse and rat models . Although ‘L’ is ~ L /D, where D is the diffusivity of the drug inside Dissociation constant, KD cartilage tissue. Here, the concentration of the drug at which (in equilibrium) rodent models continue to have an essential half of the binding sites are occupied by the drug. Generally, role in our understanding of the biological the lower the value of K the tighter the binding. Electrostatic interaction D mechanisms underlying OA and PTOA, Non-covalent repulsive or attractive interaction and therefore in initial drug screening and between charged molecules (for example, proteins, Binding site density, NT glycosaminoglycan chains) in a physiological medium Here, the local density of sites inside a tissue that can bind drug discovery, these models might not (for example, saline, synovial fluid) or inside highly drug molecules. be informative regarding drug delivery. charged tissues such as cartilage. Transport kinetics must be investigated Binding affinity using larger animal models with thicker Partition coefficient Here, the strength of the binding interaction between The equilibrium concentration of unbound, free drug inside a drug and its binding-site partner that bind together cartilage more like human, models that cartilage, normalized to drug concentration in the synovial reversibly. High affinity means very tight binding. might also be more clinically relevant (and fluid (denoted as k). generally preferred by the FDA75,76). Drug Dynamic loading delivery and transport kinetics depend Electrical potential The mechanical loading of joints, which can occur across a on drug carrier size and surface-func- The potential energy of a charged particle at any location wide range of frequencies (loading rates) depending on the divided by the particle’s charge. Sharp jumps in electrical type of physical activity. For example, joint loading tional properties, and on the biophysical potential result in high localized electric fields at that region. frequencies can range from <1Hz in slow activities such as properties of the animal joint and its walking to 1,000 Hz for high rate activities such as jumping constituent tissues. The size of joint space Steric hindrance and high impact sports. and, in particular, the thickness of cartilage When the pore size of the tissue matrix is small enough, 77,78 diffusion and transport of a drug or drug-carrier will be Cationic drug nanocarriers increase with animal size . For example, hindered simply because of its size. Biological or synthetic nano-particles (with diameters average cartilage thickness for different approximately <10 nm) that can be conjugated to small mature animal species have been reported Donnan partitioning or large molecule drugs to enhance delivery. as follows: mouse ~50µm, rat ~100–150µm, The change in concentration of a charged drug across rabbit ~350–700µm, goat ~900µm, pig the synovial fluid–cartilage interface due to the drug’s Electrostatic binding charge. The high negative fixed-charge density of Binding due to electrostatic interactions; generally ~1.5 mm, horse ~1.5– 2.0 mm and human glycosaimnoglycans inside cartilage results in a drop in the nonspecific and much weaker than strong (for example, ~1.5–2.0 mm (REFS 77–80). Therefore, electrical potential at the tissue interface, causing a strong covalent) binding. drug uptake, diffusion-reaction transport

190 | MARCH 2017 | VOLUME 13 www.nature.com/nrrheum ©2017 Mac millan Publishers Li mited, part of Spri nger Nature. All ri ghts reserved. ©2017 Mac millan Publishers Li mited, part of Spri nger Nature. All ri ghts reserved.

PERSPECTIVES

a Non-penetrating particles b Surface binding particles c Charged penetrating particles

Cartilage Cartilage Cartilage

Bone Bone Bone

kC

τ2

C C C τ1 C′ * * Concentration * Concentration τ2 Concentration τ1 Synovial fluid Cartilage Bone Synovial fluid Cartilage Bone Synovial fluid Cartilage Bone

• No penetration of drug carriers • Surface adsorption and binding of • Rapid and full-depth penetration • No binding to cartilage ECM drug carriers limits kinetics • Electrostatic interactions and • Carriers suspended in synovial fluid • Large carriers sterically hindered Donnan partitioning • Drug is rapidly cleared out thus C′< C

• Increasing effectiveness of drug delivery systems • Increasing rate of achieving drug therapeutic threshold

Figure 4 | Approaches to intra-articular drug delivery. a | Large, non-pen- by Donnan partitioning, which accelerates drug transport into cartilage etrating, non-binding drug carriers remain suspended in synovial fluid. faster than drug clearance from the synovialNature fluid. Reviews Thus, the | Rheumatology time to reach These carriers are most relevant when the target sites of the drug are intra-cartilage therapeutic levels is shortest in this case compared with mainly in the synovial fluid or synovial membrane, such as with drugs used large carriers or surface-binding carriers. A graph for each drug delivery for relieving pain and inflammation. b | Large or small carriers (depicted system depicts the transient carrier and drug concentration profiles.

in blue and orange respectively) that bind strongly to the cartilage sur- τ1 and τ2 denote an earlier and later time point respectively. C is the con- face and are unable to penetrate deeper into the tissue are relevant for centration of drug (encapsulated in carriers) in the synovial fluid. Cʹ is the the delivery of drugs to target sites at or near the tissue surface. effective drug concentration in synovial fluid after clearance from c | Carriers with optimal size and positive charge can penetrate through the joint. Blue and orange curves show the concentration gradient of the full thickness of cartilage and be retained owing to weak, reversible drug carriers. Red dotted curves show concentration gradient of the drug binding interactions. The sharp increase in cationic drug carrier concen- released from these carriers inside the cartilage. * denotes the drug tration (from C to kC) at the synovial fluid–cartilage interface is caused therapeutic threshold.

kinetics, and retention will vary markedly cartilage compared to the thinner rat drug delivery into negatively charged with animal species. Whereas drug carriers cartilage52. The longest half-life of avidin was tissues such as cartilage by either might penetrate rapidly into 50µm‑thick measured in the thickest cartilage of rabbits functionalizing drugs with cationic peptide mouse cartilage, in larger animals and (medial tibial plateau, 155 h) whereas in rat domains or utilizing cationic nanocarriers. humans they could easily be cleared from cartilage the half-life of avidin was five to By designing drug-carrier conjugates of the joints before much penetration, as the six times shorter47,52. Much longer retention optimal size and charge, it is possible to diffusion time is proportional to the square of times would be expected in thicker human enable their penetration and long-term cartilage thickness. Conversely, once a drug cartilage than in rabbits. Taken together, retention through the full thickness of reaches therapeutic levels inside cartilage, rodent models might greatly overestimate cartilage, which is necessary for drug the theoretical retention time favours thicker drug and/or drug carrier penetration into delivery to chondrocytes and other ECM cartilage. The outward diffusion-reaction time cartilage, compared to the much thicker targets. This approach could enable is also proportional to cartilage thickness human cartilage, and greatly underestimate treatment of early stage OA and PTOA squared, and inversely proportional to the drug and/or drug carrier retention. when the disease is still ‘reversible’. In the effective diffusivity of the drug and/or case of late-stage OA, this approach might drug carrier inside cartilage, including Conclusions enhance the residence times of symptomatic the effects of binding. For example, avidin Electrostatic (charge–charge) interactions medication by enabling binding within the takes much longer to diffuse out of rabbit provide a unique opportunity for targeted synovial joint.

NATURE REVIEWS | RHEUMATOLOGY VOLUME 13 | MARCH 2017 | 191 ©2017 Mac millan Publishers Li mited, part of Spri nger Nature. All ri ghts reserved. ©2017 Mac millan Publishers Li mited, part of Spri nger Nature. All ri ghts reserved.

PERSPECTIVES

Ambika G. Bajpayee is at the Department of 20. Shamji, M. F. et al. Development and characterization glycosaminoglycans visualized via atomic force Bioengineering, Northeastern University, of a fusion protein between thermally responsive microscopy. J. Struct. Biol. 143, 242–257 (2003). elastin-like polypeptide and interleukin‑1 receptor 46. Buschmann, M. D. & Grodzinsky, A. J. A molecular 360 Huntington Avenue, Boston, Massachusetts antagonist: sustained release of a local model of proteoglycan-associated electrostatic forces 02115, USA. antiinflammatory therapeutic. Arthritis Rheum. 56, in cartilage mechanics. J. Biomech. Eng. 117, Alan J. Grodzinsky is at the Departments of Biological 3650–3661 (2007). 179–192 (1995). 21. Meinel, L. & Kaplan, D. L. Silk constructs for delivery 47. Bajpayee, A. G., Scheu, M., Grodzinsky, A. J. & Engineering, Mechanical Engineering, and Electrical of muskuloskeletal therapeutics. Adv. Drug Deliv. Rev. Porter, R. M. Electrostatic interactions enable rapid Engineering and Computer Science, Massachusetts 64, 1111–1122 (2012). penetration, enhanced uptake and retention of intra- Institute of Technology, 77 Massachusetts Ave, 22. Zeng, J. et al. Biodegradable electrospun fibers for drug articular injected avidin in rat knee joints. J. Orthop. Cambridge, Massachusetts 02139, USA. delivery. J. Control. Release 92, 227–231 (2003). Res. 32, 1044–1051 (2014). 23. Heinegård, D. & Saxne, T. The role of the cartilage 48. Sterner, B. et al. The effect of polymer size and charge of Correspondence to A.J.G. matrix in osteoarthritis. Nat. Rev. Rheumatol. 7, molecules on permeation through synovial membrane [email protected] 50–56 (2011). and accumulation in hyaline articular cartilage. 24. Bajpayee, A. G., Wong, C. R., Bawendi, M. G., Eur. J. Pharm. Biopharm. 101, 126–136 (2016). doi:10.1038/nrrheum.2016.210 Frank, E. H. & Grodzinsky, A. J. Avidin as a model for 49. Snowden, J. M. & Maroudas, A. The distribution of Published online 9 Feb 2017 charge driven transport into cartilage and drug serum albumin in human normal and degenerate delivery for treating early stage post-traumatic articular cartilage. Biochim. Biophys. Acta 428, osteoarthritis. Biomaterials 35, 538–549 (2014). 726–740 (1976). 1. Matthews, G. L. & Hunter, D. J. Emerging drugs for 25. Wieland, H. A., Michaelis, M., Kirschbaum, B. J. & 50. Grodzinsky, A. J. Fields, Forces, and Flows in osteoarthritis. Expert Opin. Emerg. Drugs 16, 479–491 Rudolphi, K. A. Osteoarthritis — an untreatable Biological Systems 89–99 (Garland Science, 2011). (2011). disease? Nat. Rev. Drug Discov. 4, 331–344 51. Maroudas, A. Transport of solutes through cartilage: 2. Gerwin, N., Hops, C. & Lucke, A. Intraarticular drug (2005). permeability to large molecules. J. Anat. 122, delivery in osteoarthritis. Adv. Drug Deliv. Rev. 58, 26. Huebner, K. D., Shrive, N. G. & Frank, C. B. 335–347 (1976). 226–242 (2006). Dexamethasone inhibits inflammation and cartilage 52. Bajpayee, A. G., Scheu, M., Grodzinsky, A. J. & 3. Evans, C. H., Kraus, V. B. & Setton, L. A. Progress in intra- damage in a new model of post-traumatic Porter, R. M. A rabbit model demonstrates the articular therapy. Nat. Rev. Rheumatol. 10, 11–12 osteoarthritis. J. Orthop. Res. 32, 566–572 (2014). influence of cartilage thickness on intra-articular drug (2014). 27. Grodzinsky, A. J., Wang, Y., Kakar, S., Vrahas, M. S. & delivery and retention within cartilage. J. Orthop. Res. 4. Larsen, C. et al. Intra-articular depot formulation Evans, C. H. Intra-articular dexamethasone to inhibit 33, 660–667 (2015). principles: role in the management of postoperative pain the development of post-traumatic osteoarthritis. 53. Hughes, C. et al. Human single-chain variable and arthritic disorders. J. Pharm. Sci. 97, 4622–4654 J. Orthop. Res. http://dx.doi.org/10.1002/jor.23295 fragment that specifically targets arthritic cartilage. (2008). (2016). Arthritis Rheum. 62, 1007–1016 (2010). 5. Evans, C. H. Drug delivery to chondrocytes. Osteoarthritis 28. Chevalier, X. et al. Safety study of intraarticular 54. Byun, S. et al. Transport and equilibrium uptake of a Cartilage 24, 1–3 (2016). injection of interleukin 1 receptor antagonist in peptide inhibitor of PACE4 into articular cartilage is 6. Anderson, D. D. et al. Post-traumatic osteoarthritis: patients with painful knee osteoarthritis: a multicenter dominated by electrostatic interactions. Arch. improved understanding and opportunities for early study. J. Rheumatol. 32, 1317–1323 (2005). Biochem. Biophys. 499, 32–39 (2010). intervention. J. Orthop. Res. 29, 802–809 (2011). 29. Kraus, V. B. et al. Effects of intraarticular IL1‑Ra for 55. Nia, H. T. et al. High-bandwidth AFM-based rheology 7. Johnson, D. L., Urban, W. P., Caborn, D. N. M., acute anterior cruciate ligament knee injury: a reveals that cartilage is most sensitive to high loading Vanarthos, W. J. & Carlson, C. S. Articular cartilage randomized controlled pilot trial (NCT00332254). rates at early stages of impairment. Biophys. J. 104, changes seen with magnetic resonance imaging-detected Osteoarthritis Cartilage 20, 271–278 (2012). 1529–1537 (2013). bone bruises associated with acute anterior cruciate 30. Loffredo, F. S. et al. Targeted delivery to cartilage is 56. O’Hara, B. P., Urban, J. P. & Maroudas, A. Influence of ligament rupture. Am. J. Sports Med. 26, 409–414 critical for in vivo efficacy of insulin-like growth factor 1 cyclic loading on the nutrition of articular cartilage. (1998). in a rat model of osteoarthritis. Arthritis Rheumatol. Ann. Rheum. Dis. 49, 536–539 (1990). 8. Rothenfluh, D. A., Bermudez, H., O’Neil, C. P. & 66, 1247–1255 (2014). 57. Evans, R. C. & Quinn, T. M. Dynamic compression Hubbell, J. A. Biofunctional polymer nanoparticles for 31. Miller, R. E. et al. Intraarticular injection of heparin- augments interstitial transport of a glucose-like solute in intra-articular targeting and retention in cartilage. binding insulin-like growth factor 1 sustains delivery of articular cartilage. Biophys. J. 91, 1541–1547 (2006). Nat. Mater. 7, 248–254 (2008). insulin-like growth factor 1 to cartilage through 58. Bonassar, L. J. et al. The effect of dynamic 9. Morgen, M. et al. Nanoparticles for improved local binding to chondroitin sulfate. Arthritis Rheum. 62, compression on the response of articular cartilage to retention after intra-articular injection into the knee joint. 3686–3694 (2010). insulin-like growth factor‑I. J. Orthop. Res. 19, 11–17 Pharm. Res. 30, 257–268 (2013). 32. Onuora, S. Osteoarthritis: sprifermin shows cartilage- (2001). 10. Kang, M. L., Ko, J.‑Y., Kim, J. E. & Im, G.‑I. Intra-articular protective effects in knee OA. Nat. Rev. Rheumatol. 59. Petrella, R. J. et al. Safety and performance of Hydros delivery of kartogenin-conjugated chitosan nano/ 10, 322–322 (2014). and Hydros‑TA for knee osteoarthritis: a prospective, microparticles for cartilage regeneration. Biomaterials 33. US National Library of Medicine. ClinicalTrials.gov multicenter, randomized, double-blind feasibility trial. 35, 9984–9994 (2014). https://clinicaltrials.gov/ct2/show/NCT01033994 BMC Musculoskelet. Disord. 16, 57 (2015). 11. Cheng, C. J., Tietjen, G. T., Saucier-Sawyer, J. K. & (2014). 60. Kimmerling, K. A. et al. Sustained intra-articular Saltzman, W. M. A holistic approach to targeting disease 34. Hunter, D. J. et al. Phase 1 safety and tolerability delivery of IL‑1Ra from a thermally responsive elastin with polymeric nanoparticles. Nat. Rev. Drug Discov. 14, study of BMP‑7 in symptomatic knee osteoarthritis. like polypeptide as a therapy for post-traumatic 239–247 (2015). BMC Musculoskelet. Disord. 11, 232 (2010). arthritis. Eur. Cell. Mater. 29, 124–140 (2015). 12. Bodick, N. et al. An intra-articular, extended-release 35. Johnson, K. et al. A stem cell–based approach to 61. Vugmeyster, Y. et al. Disposition of human formulation of triamcinolone acetonide prolongs and cartilage repair. Science 336, 717–721 (2012). recombinant lubricin in naive rats and in a rat model amplifies analgesic effect in patients with osteoarthritis of 36. Hunter, D. J. Pharmacologic therapy for osteoarthritis of post-traumatic arthritis after intra-articular or the knee. J. Bone Joint Surg. Am. 97, 877–888 (2015). — the era of disease modification. Nat. Rev. intravenous administration. AAPS J. 14, 97–104 13. Zhu, S. et al. Inhibition of Rac1 activity by controlled Rheumatol. 7, 13–22 (2011). (2012). release of NSC23766 from chitosan microspheres 37. Byun, S. et al. Transport of anti‑IL‑6 antigen binding 62. Sacchetti, C. et al. Polyethylene-glycol-modified single- effectively ameliorates osteoarthritis development in vivo. fragments into cartilage and the effects of injury. walled carbon nanotubes for intra-articular delivery to Ann. Rheum. Dis. 74, 285–293 (2013). Arch. Biochem. Biophys. 532, 15–22 (2013). chondrocytes. ACS Nano 8, 12280–12291 (2014). 14. Dong, J. et al. Intra-articular delivery of liposomal 38. Chevalier, X. et al. Intraarticular injection of anakinra 63. Bajpayee, A. G., Quadir, M. A., Hammond, P. T. & celecoxib–hyaluronate combination for the treatment of in osteoarthritis of the knee: a multicenter, Grodzinsky, A. J. Charge based intra-cartilage delivery osteoarthritis in rabbit model. Int. J. Pharm. 441, randomized, double-blind, placebo-controlled study. of single dose dexamethasone using Avidin nano- 285–290 (2013). Arthritis Care Res. 61, 344–352 (2009). carriers suppresses cytokine-induced catabolism long 15. Urbano, P. C. et al. Effect of pegylated 39. Cohen, S. et al. A randomized, double-blind study of term. Osteoarthritis Cartilage 24, 71–81 (2016). phosphatidylserine-containing liposomes in experimental AMG 108 (a fully human monoclonal antibody to 64. Hu, H.‑Y. et al. In vivo visualization of osteoarthritic chronic arthritis. BMC Pharmacol. Toxicol. 16, 24 (2015). IL‑1R1) in patients with osteoarthritis of the knee. hypertrophic lesions. Chem. Sci. 6, 6256–6261 16. Vemula, P. K. et al. On‑demand drug delivery from self- Arthritis Res. Ther. 13, R125 (2011). (2015). assembled nanofibrous gels: a new approach for 40. Apparailly, F. & Jorgensen, C. siRNA-based therapeutic 65. Hellio Le Graverand-Gastineau, M.‑P. OA clinical trials: treatment of proteolytic disease. J. Biomed. Mater. Res. A approaches for rheumatic diseases. Nat. Rev. current targets and trials for OA. Choosing molecular 97, 103–110 (2011). Rheumatol. 9, 56–62 (2013). targets: what have we learned and where we are 17. Petit, A. et al. Sustained intra-articular release of 41. Pi, Y. et al. Intra-articular delivery of anti-Hif‑2α siRNA headed? Osteoarthritis Cartilage 17, 1393–1401 celecoxib from in situ forming gels made of acetyl-capped by chondrocyte-homing nanoparticles to prevent (2009). PCLA-PEG-PCLA triblock copolymers in horses. cartilage degeneration in arthritic mice. Gene Ther. 66. Gouze, J.‑N. et al. Adenovirus-mediated gene transfer Biomaterials 53, 426–436 (2015). 22, 439–448 (2015). of glutamine: fructose-6‑phosphate amidotransferase 18. Réeff, J. et al. Development and evaluation in vitro and 42. Maroudas, A. in Adult Articular Cartilage (ed. antagonizes the effects of interleukin‑1β on rat in vivo of injectable hydrolipidic gels with sustained- Freeman, M. A. R.) 215–290 (Kent, 1979). chondrocytes. Osteoarthritis Cartilage 12, 217–224 release properties for the management of articular 43. Eyre, D. R. & Weis, M. A. & Wu, J. -J. Articular (2004). pathologies such as osteoarthritis. Int. J. Pharm. 490, cartilage collagen: an irreplaceable framework. 67. Nieminen, H. J. et al. Delivering agents locally into 74–84 (2015). Eur. Cell. Mater. 12, 57–63 (2006). articular cartilage by intense MHz ultrasound. 19. Matsuzaki, T. et al. Intra-articular administration of 44. Comper, W. D. Cartilage: Molecular Aspects Ch.2 Ultrasound Med. Biol. 41, 2259–2265 (2015). gelatin hydrogels incorporating rapamycin-micelles (ed. Hall, B. K. & Newman, S. A.) 59–96 (CRC 68. Coombe, D. R. Biological implications of reduces the development of experimental osteoarthritis Press,1991). glycosaminoglycan interactions with haemopoietic in a murine model. Biomaterials 35, 9904–9911 45. Ng, L. et al. Individual cartilage aggrecan cytokines. Immunol. Cell Biol. 86, 598–607 (2014). macromolecules and their constituent (2008).

192 | MARCH 2017 | VOLUME 13 www.nature.com/nrrheum ©2017 Mac millan Publishers Li mited, part of Spri nger Nature. All ri ghts reserved. ©2017 Mac millan Publishers Li mited, part of Spri nger Nature. All ri ghts reserved.

PERSPECTIVES

69. Haglund, L., Tillgren, V., Önnerfjord, P. & 76. Gregory, M. H. et al. A review of translational animal 82. American Academy of Orthopaedic Surgeons. Heinegård, D. The C‑terminal peptide of models for knee osteoarthritis. Arthritis 2012, Treatment of osteoarthritis of the knee. Evidence- chondroadherin modulates cellular activity by e764621 (2012). based guideline (American Academy of Orthopedic selectively binding to heparan sulfate chains. 77. Malda, J. et al. Of mice, men and elephants: the Surgeons, 2013). J. Biol. Chem. 288, 995–1008 (2013). relation between articular cartilage thickness and body 70. Rucci, N. et al. The glycosaminoglycan-binding domain mass. PLoS ONE 8, e57683 (2013). Acknowledgements of PRELP acts as a cell type-specific NF‑κB inhibitor 78. Kamisan, N., Naveen, S. V., Ahmad, R. E. & Tunku, K. This work was supported by NIH National Institute of that impairs osteoclastogenesis. J. Cell Biol. 187, Chondrocyte density, proteoglycan content and gene Biomedical Imaging and Bioengineering grant EB017755, 669–683 (2009). expressions from native cartilage are species specific National Science Foundation Materials Research Science and 71. Pratta, M. A. et al. Aggrecan protects cartilage and not dependent on cartilage thickness: a Engineering Centers (MRSEC) grant DMR‑1419807, NIH collagen from proteolytic cleavage. J. Biol. Chem. 278, comparative analysis between rat, rabbit and goat. National Institute of Arthritis and Musculoskeletal and Skin 45539–45545 (2003). BMC Vet. Res. 9, 62 (2013). Diseases (NIAMS) grant AR060331 and Department of 72. Li, Y. et al. Effects of insulin-like growth factor‑1 and 79. Hoch, D. H., Grodzinsky, A. J., Koob, T. J., Albert, M. L. Defense (DoD) Congressionally Directed Medical Research dexamethasone on cytokine-challenged cartilage: & Eyre, D. R. Early changes in material properties of Programs (CDMRP) grant W81XWH-14-1-0544. relevance to post-traumatic osteoarthritis. rabbit articular cartilage after meniscectomy. J. Orthop. Osteoarthritis Cartilage 23, 266–274 (2015). Res. 1, 4–12 (1983). Author contributions 73. Malfait, A.‑M. & Little, C. B. On the predictive utility of 80. Frisbie, D. D., Cross, M. W. & McIlwraith, C. W. A Both authors researched data for the article, provided sub- animal models of osteoarthritis. Arthritis Res. Ther. comparative study of articular cartilage thickness in the stantial contributions to discussion of its content, wrote the 17, 225 (2015). stifle of animal species used in human pre-clinical studies article and undertook review and/or editing of the manuscript 74. Hunter, D. J. & Little, C. B. The great debate: compared to articular cartilage thickness in the human before submission. should osteoarthritis research focus on ‘mice’ or knee. Vet. Comp. Orthop. Traumatol. 19, 142–146 (2006). ‘men’? Osteoarthritis Cartilage 24, 4–8 (2016). 81. Nia, H. T. et al. High-bandwidth AFM-based rheology Competing interests statement 75. Little, C. B. & Hunter, D. J. Post-traumatic reveals that cartilage is most sensitive to high loading A.G.B and A.J.G. declare that they are inventors on the US osteoarthritis: from mouse models to clinical trials. rates at early stages of impairment. Biophys. J. 104, patent 9289506 B2 ‘Surface binding of nanoparticles-based Nat. Rev. Rheumatol. 9, 485–497 (2013). 1529–1537 (2013). drug delivery to tissue’.

NATURE REVIEWS | RHEUMATOLOGY VOLUME 13 | MARCH 2017 | 193 ©2017 Mac millan Publishers Li mited, part of Spri nger Nature. All ri ghts reserved. ©2017 Mac millan Publishers Li mited, part of Spri nger Nature. All ri ghts reserved.