Targeting the extracellular matrix for delivery of bioactive molecules to sites of arthritis

Abstract

Modifications to the extracellular matrix (ECM) can be either causal or consequential of disease processes including arthritis and cancer. In arthritis, the cartilage ECM is adversely affected by the aberrant behaviours of inflammatory cells, synoviocytes and chondrocytes, which secrete a plethora of cytokines and degradative proteases. In cancer, the ECM and stromal cells are linked to disease severity, and metalloproteinases are implicated in metastasis. There have been some successes in the field of targeted therapies, but efficacy depends upon the type and stage of disease. ECM targets are becoming increasingly attractive for drug delivery, owing to changes in ECM structure and composition in the diseased state, and the long in vivo half‐life of its components. This review will highlight various strategies for targeting therapeutics to arthritic joints, including antibody and peptide‐mediated drug delivery platforms to aid delivery to the ECM and retention at disease sites.

Linked Articles

This article is part of a themed section on Translating the Matrix. To view the other articles in this section visit http://onlinelibrary.wiley.com/doi/10.1111/bph.v176.1/issuetoc

Abbreviations

AA

adjuvant‐induced arthritis

ADC

antibody‐drug conjugate

ADME

absorption, distribution, metabolism and excretion

AGG

aggrecanase

BMP

bone morphogenetic protein

CAF

cancer‐associated fibroblast

CIA

collagen‐induced arthritis

COL2A1

type‐II collagen

DMARD

disease‐modifying anti‐rheumatic drug

ECM

extracellular matrix

EF

oedema factor

EPR

enhanced permeability and retention

GAG

glycosaminoglycan

GC

glucocorticoid

GF

growth factor

HPMA

poly(N‐(2‐hydroxypropyl)methacrylamide‐lactate)

ICAM

intercellular adhesion molecule

IGF

insulin‐like growth factor

LAP

latency‐associated peptide

LF

lethal factor

MSC

mesenchymal stem cell

MSH

melanocyte‐stimulating hormone

NBD

NEMO‐binding domain

NSAID

non‐steroidal anti‐inflammatory drug

OA

osteoarthritis

PA

protective antigen

PDC

peptide‐drug conjugate

PEG

poly(ethylene glycol)

PPS

poly(propylene sulphide)

PTD

protein transduction domain

RA

rheumatoid arthritis

scFv

single chain Fv

SCID

severe combined immunodeficiency

syETP

synovial endothelial targeting peptide

TenC

tenascin‐C

VIP

vasoactive intestinal peptide

Introduction

The extracellular matrix (ECM) is a complex network of proteins affecting cell shape, tissue architecture and organ function. Its precise composition and properties are highly tissue specific. The ECM is dynamic, with the speed of tissue turnover governed by both the tissue/organ type and the state of health or disease in the micro‐environment. The healthy ECM contains a reservoir of growth factors (GFs), and local cell behaviour is delicately orchestrated in a spatio‐temporal, pH and enzymatically regulated manner. In disease, changes to the ECM can be either causal or consequential. In arthritic diseases including osteoarthritis (OA) and rheumatoid arthritis (RA), the cartilage and bone ECM becomes damaged and repair processes become erroneous. This contributes to the aberrant behaviours of chondrocytes, synovial fibroblasts and immune cells (Smolen and Steiner, 2003). Moreover, the degraded ECM can actively cause further tissue deterioration in RA, with the generation of cartilage neo‐epitopes initiating and potentiating autoimmune disease (van Beers et al., 2012; Strollo et al., 2013).

OA and RA are a significant global problem. There is no cure for either disease, and treatments give mixed results. They are typically treated with non‐steroidal anti‐inflammatory drugs (NSAIDs), glucocorticoids (GCs), small molecule disease‐modifying anti‐rheumatic drugs (DMARDs) and biological DMARDs. Joint replacement surgery is used in extreme cases but is costly to health services, presents high risks and there are no guarantees of long‐term patient remission. Treatments only address the pain and inflammation, and currently, there are none that can stimulate regeneration of the damaged cartilage and bone. NSAIDs are mainly used for palliative purposes, although they do alleviate a degree of disease‐associated inflammation. GCs and DMARDs such as methotrexate reduce inflammation but are only moderately efficacious. The modest therapeutic effects of NSAIDs, GCs and DMARDs can be attributed to their poor bioavailability at sites of disease. Additionally, systemically administered agents with ubiquitous therapeutic targets can cause side effects. Importantly, long‐term GC usage is associated with osteoporosis, limiting their use (Smolen and Steiner, 2003; McDonough et al., 2008).

Targeted therapies that only function in diseased tissues have multiple benefits over free drug molecules. They permit the administration of lower doses, reducing the risk of side effects and leading to an improved drug safety profile. The need for concomitant therapies to treat side effects can therefore be negated. Targeted drugs may also increase the safety of delivering a higher, more efficacious dose since peripheral tissues are less affected. Finally, longer remission periods may be possible owing to an increased drug half‐life. Targeting drugs to the ECM is an increasingly popular therapeutic strategy for arthritis for several reasons. ECM molecules of the joint have long half‐lives, widening the therapeutic window for their targeting. Furthermore, the efflux of drugs from the synovium can be rapid and occur before a sufficient therapeutic effect has been achieved (Evans et al., 2013), so their anchorage to the ECM may achieve greater therapeutic efficacy due to longer residency at disease sites. Modifications to the ECM through tissue remodelling or angiogenesis can be utilized for targeting (Lee et al., 2002; Trachsel et al., 2007; Hughes et al., 2010).

Systemically administered, locally acting therapies are increasingly seen in the clinic, but their ultimate success depends on further developments. This review will firstly give an overview of some of the methods used to treat OA and RA – both diseases with ECM involvement. Secondly, it will describe ways in which the ECM itself is increasingly being used to aid drug delivery and retention. Some technologies directly engage components of the ECM and the cells accountable for its maintenance, whilst others are preferentially retained in the ECM or are activated by local disease processes. Appropriate drug encapsulation together with targeting technologies allow ECM site‐specific therapeutic delivery.

Extending drug half‐life to increase efficacy and targeting

The absorption, distribution, metabolism and excretion (ADME) of therapeutics are governed by molecular size, charge and post‐translational modifications (for protein therapeutics). The site or route of administration is also critical, with the ECM largely influential on entry into the circulation, or tissue admission and retention (Tibbitts et al., 2016). Once in the circulation, the enhanced permeability of fenestrated neovasculature allows drugs of an appropriate size to enter and remain within tissues, whereas usually they would be retained in the circulation (Maeda et al., 2013). This enhanced permeability and retention (EPR) assists the passive therapeutic targeting to RA joints. This section will briefly outline several key technologies affecting ADME and EPR.

Polymers can be used to alter drug stability and release kinetics. Poly(ethylene glycol) (PEG) enhances the in vivo half‐life of coupled payloads by increasing their molecular weight, and reduces off‐target drug uptake. PEGylated certolizumab (Figure 1C) is a licenced anti‐TNFα antibody fragment used to treat RA (Curtis et al., 2015). Additionally, capsules of the biodegradable poly(lactic‐co‐glycolic acid; PLGA) containing triamcinolone acetonice (TCA) are in development to treat OA (Kumar et al., 2015), since they improve steroid delivery and release. The polymer HPMA has also been conjugated to the steroid dexamethasone in both fast‐release and slow‐release formulations, with the slower‐release derivative being more efficacious in an adjuvant‐induced arthritis (AA) model (Quan et al., 2014). Various amphiphilic subunits (Figure 1A) can be used to package drugs into micelles or liposomes, which are also useful for regulating the release of hydrophobic drugs like dexamethasone for treating arthritis. By coupling the steroid to a PEG/HPMA amphiphile via ester bonds (Figure 1B), micelles were produced with improved efficacy when compared to the free drug in preclinical models. Differing degrees of release were achievable by using either sulphide, sulphoxide or sulphone ester linkages between drug and amphiphile (Crielaard et al., 2012). Moreover, PEG/poly(ε‐caprilactone) amphiphilic micelles loaded with low‐dose dexamethasone were highly efficacious in a model of AA (Wang et al., 2016). Liposomes are loaded with either hydrophilic or hydrophobic therapeutics, and liposomal dexamethasone (Figure 1D) has also been shown to ameliorate adjuvant‐induced arthritis (AA; Quan et al., 2014). These examples demonstrate the diverse effects on drug delivery that can be achieved by varying the formulation, allowing the manipulation of drug distribution, circulation time, release kinetics and dose.

Figure 1.

Drug modifications that extend their half‐lives. Schematics are shown here of a generic amphiphilic subunit (A) used to incorporate dexamethasone in micelle (B) or liposomal (D) formulations. PEG is coupled to RA drugs such as the biological agent certolizumab pegol (C), thereby increasing circulation time and improving therapeutic activity.

Active drug targeting technologies

The cartilage and sub‐endothelial ECM are increasingly being utilized for the development of targeted drugs for OA and RA. Briefly outlined here are the overarching methods used for targeted therapeutic delivery, with examples of ECM targeting versions to follow.

Antibodies

Antibodies have become the workhorse of both the laboratory and the clinic, and many are now FDA‐approved drugs. Effector function is dictated by their constant region (Fc), and the variable region (Fv) binds designated target antigens (Figure 2A). Therapeutically, antibodies neutralize, block or label their antigens for elimination by antibody‐dependent cellular cytotoxicity or complement‐dependent cytotoxicity. Antibodies have revolutionized RA management through blockade of soluble pro‐inflammatory factors such as the cytokine TNFα (Elliott et al., 1994).

Figure 2.

Targeting agents to facilitate disease‐specific drug accumulation. Schematics are shown of various constituent fragments derived from engineering an antibody including scFv, diabodies and Fab fragments (A). Antibodies and derivative fragments can be coupled to drug molecules including cytokines as antibody‐drug conjugates (B), and synthetic targeting peptides can similarly be linked to therapeutics as peptide‐drug conjugates (C).

Antibody fragments

The modular structure of antibodies permits the engineering of antibody‐based molecules such as the monovalent single chain Fv (scFv) and antibody‐related fusion proteins including immunocytokines (an antibody component fused to a therapeutic cytokine). ScFv contains the two hypervariable antigen‐binding fragments of an antibody linked by a glycine‐serine linker region. Its smaller size allows more efficient tissue penetration than a full antibody. Antibody fragments take various additional forms (Figure 2A) including diabodies (two scFv fragments fused with a linker peptide) and Fab fragments (scFv plus the CH1 domain of the heavy chain and CL of the light chain) (Young et al., 2014).

Antibody‐drug conjugates

Since antibodies can localize to specific tissues, antibody‐drug conjugates (ADCs) (Figure 2B) could be employed to facilitate the delivery and reduce off‐target effects of toxic agents in arthritis. Lessons can be learned from the oncology field, where tissue‐specific targeting and retention of the chemotherapeutic agent emtansine was achieved when conjugated to trastuzumab (Herceptin). This compound (T‐DM1) is clinically approved for late‐stage breast cancer patients, but emtansine had previously failed clinical trials due to being too toxic when untargeted (Verma et al., 2012).

Cytokines are potent immune regulators with huge therapeutic potential, but there are issues with their application for treating arthritis. These include the ubiquitous expression of their receptors leading to systemic toxic effects and their short half‐lives causing bioavailability at diseased tissues to be below the efficacious threshold. Several cytokines have been dropped from clinical trials for these reasons (Keystone et al., 1998; Morgan et al., 2012). Immunocytokines have been employed to enhance cytokine concentrations at disease sites and reduce effects in peripheral tissues. Fusions of the pro‐inflammatory cytokine IL‐2 to antibodies recognizing tumour‐associated dinutuximab or rituximab have shown efficacy during clinical development for melanoma and non‐Hodgkin's lymphoma respectively (Young et al., 2014). For RA, ADCs of anti‐inflammatory cytokines such as IL‐4 and IL‐10 are also being developed, which are discussed later.

Peptide‐drug conjugates

Short peptides with affinity for ECM have been isolated by phage‐display and can be conjugated to biologically active molecules (Figure 2C) to facilitate their targeting (Rothenfluh et al., 2008; Cheung et al., 2013; Wythe et al., 2013). Polyvalent molecules containing multiple copies of targeting peptides may remain bound to a target longer or more strongly than monovalent equivalents, aiding tissue retention (Wythe et al., 2013).

Matrix activated prodrugs

Pathology‐related, enzymatically mediated ECM remodelling can be utilized for drug targeting. MMPs are elevated in the synovium of OA patients and even more so in RA (Yoshihara et al., 2000). To this end, MMP‐activated prodrugs have been developed (Figure 3A) for RA incorporating antibodies and cytokines, and GFs for OA.

Figure 3.

Matrix‐specific enzymatically activated prodrugs to facilitate targeted therapeutic delivery. Schematics are depicted of various prodrug technologies which are activated by disease‐associated protease activity. LAP‐fusions to a payload (cytokine, GF or peptide) via an MMP or AGG site (A) confers latency to the cargo molecule until reaching sites rich in these enzymes, where LAP is cleaved and steric hindrance is attenuated. Activatable dual‐variable domain (aDvD) antibodies have been developed with an initial ICAM‐1 binding capacity, which subsequently sequester TNFα following MMP cleavage (B). C‐terminal PEGylation of an MMP‐7 cleavable peptide followed by N‐terminal coupling to a cationic polymeric motif encapsulates siRNA until enzymatic cleavage and exposure of the positively charged core (C), leading to cell delivery.

Disease‐specific ECM features for targeted therapies

Differentiating between the disease site and healthy peripheral tissue is critical for developing a targeted therapy. Changes in tissue composition or physiological state of the micro‐environment can be utilized for this purpose.

Targeting articular cartilage in arthritis

Cartilage is avascular and so cannot be targeted via the circulation; however, drug retention following local intra‐articular delivery can be achieved, which may improve the therapeutic indices of existing drugs. Synthetic peptides have been identified that interact with ECM components of the cartilage, which can aid the retention of conjugated protein payloads within the cartilage matrix, targeting peptides are listed in Table 1. The aggrecan‐binding peptides RLDPTSYLRTFW and HDSQLEALIKFM have been identified and may soon be employed for such applications (Cheung et al., 2013). The heparin‐binding peptide KRKKKGKGLGKKRDPSLRKYK was shown to be retained in the cartilage in a rat model of OA, and when fused to insulin‐like growth factor 1 (IGF‐1) reduced disease severity (Loffredo et al., 2014). Interestingly, it was shown that the tetrapodal compound DOTAM used in conjunction with the COL2A1 (type‐II collagen)‐targeting peptide WYRGRL facilitated articular cartilage localization and retention. Cathepsin D‐mediated aggrecan damage and glycosaminoglycan (GAG) release could be inhibited by DOTAM when covalently linked to pepstatin A with either one or three copies of WYRGRL. Trivalent molecules were retained more strongly than monovalent versions (Hu et al., 2015). Theoretically, loading DOTAM with other targeting technologies and payloads could achieve multiple additional novel therapeutic effects in the cartilage.

Table 1.

Peptide sequences and their identified binding partners

Peptide sequence	Binding partner	Described applications	References
DWRVIIPPRPSA	Human/rabbit chondrocytes	Plasmid DNA delivery	(Pi et al., 2011)
RLDPTSYLRTFW HDSQLEALIKFM	Aggrecan	N/A	(Cheung et al., 2013)
WYRGRL	Type‐II collagen (α1 chain)	Pepstatin A delivery (cathepsin D inhibition)	(Hu et al., 2015)
CKSTHDRLC	Human synovial endothelium	IL‐4 delivery to arthritic tissue	(Wythe et al., 2013)
CLDNQRPKC CDCRGDCFC	Vasculature of inflamed tissue	Inhibition of angiogenesis	(Yang et al., 2011)
CRNADKFPC	Vasculature of inflamed tissue	N/A	(Yang et al., 2011)
FHKHKSPALSPVGGG	Tenascin‐C	Targeting of tumour cell xenografts in vivo	(Kim et al., 2012)
KRKKKGKGLGKKRDPSLRKYK	Heparin	Retention of IGF‐1 in OA model	(Loffredo et al., 2014)
CARSKNKDC	HSPGs	Blockade of fibrosis	(Järvinen and Ruoslahti, 2010)

Current applications are listed, but these could be expanded by coupling with new payloads and encapsulation approaches.

Gene delivery is an emerging approach to treat RA and OA (Pi et al., 2011). It permits a sustained effect since more of the therapeutic gene is produced to replace that leaving the joint. Delivery of anti‐inflammatory or regenerative GF genes is a desirable gene therapy approach, since this may prevent further joint damage and facilitate cartilage anabolism (Evans et al., 2010). Non‐viral delivery of DNA to chondrocytes in vivo has been demonstrated by modifying the polycation polyethylenimine with the peptide DWRVIIPPRPSA, which is internalized specifically by chondrocytes (Pi et al., 2011). DNA delivery was demonstrated using reporter constructs but could plausibly be used to deliver therapeutic genes, with fewer peripheral cells affected.

Methods for facilitating deep ECM penetration

Technologies are also in development to aid cartilage penetration for increasing retention and to modulate chondrocyte behaviours in situ – a notoriously difficult feat (Evans et al., 2010). Nanoplexes of alternating layers of anionic poly (glutamic acid) and cationic poly (arginine) were readily able to penetrate cartilage, and IGF‐1 from loaded nanoplexes was detected in the synovium 4 weeks after intra‐articular injection compared to 9 days for free IGF‐1. IGF‐1 nanoplexes ensured cartilage integrity was retained in a rat model of cartilage injury, and synovial IL‐1β levels were significantly reduced in the nanoplex‐treated animals (Shah et al., 2016). Cartilage can also be readily penetrated by the strongly basic protein avidin due to its electrostatic attraction to negatively charged glycosaminoglycans (GAGs). Avidin has been coupled to dexamethasone, and the conjugate resided within the cartilaginous ECM and prevented IL‐1 driven GAG release in organ culture (Bajpayee et al., 2016). Such technologies could also be employed to deliver therapeutics to other ECM‐rich tissues where target cells are difficult to reach.

These examples suggest that both anti‐inflammatory and anabolic effects can be achieved in the cartilage in vivo. Together, this section demonstrates how cartilage retention and penetration may be achieved to facilitate the delivery of drugs, GFs and nucleic acids to chondrocytes in arthritic joints, although much work is still needed in this area.

Increasing efficacy of GFs by targeting heparan sulphate proteoglycans

Heparan sulphate proteoglycans (HSPGs) are ECM molecules that can interact with many growth factors to facilitate their local retention and present them to their cell surface receptors. High‐dose requirements have hampered the clinical translation of GFs, but ECM components such as HSPGs could be used to enrich local GF levels and circumvent this issue. Such benefits were conferred onto VEGFA and bone morphogenetic protein (BMP)‐2, by using a high‐affinity ECM interacting peptide from placenta GF2 with heparin binding properties. This C‐terminal peptide permitted therapeutic effects using reduced GF dosage, without the side effects of excessive GF delivery such as vascular leakage in the case of VEGF (Martino et al., 2014). Another HSPG‐interacting peptide, CARSKNKDC, has been successfully employed to deliver the TGFβ antagonist, decorin, directly to wound sites. This was shown to both enhance the ability of decorin to inhibit fibrosis and aid wound healing (Järvinen and Ruoslahti, 2010). Such approaches utilizing HSPGs could be adapted to enhance the efficacy of therapeutic GFs in arthritic disease.

Targeting RA‐associated angiogenesis

Angiogenesis facilitates the recruitment of myeloid and lymphoid cells that drive and potentiate RA, and the neovascular endothelium and its sub‐endothelial ECM has a unique molecular signature compared to established vasculature. This has led to the development of peptide‐drug conjugates (PDCs) to target drugs to these disease‐specific sites. The sequence CKSTHDRLC [referred to as synovial endothelial targeting peptide (syETP)] was shown in severe combined immunodeficiency (SCID) mice to home specifically to human OA and RA synovial xenografts (Lee et al., 2002). When fused to the cytokine IL‐4 via an MMP‐cleavable site, syETP delivered the payload to arthritic xenografts in SCID animals to elicit a shift towards an anti‐inflammatory T_H2 phenotype. Additionally, tissue retention was improved by increasing the number of syETP motifs from one to three (Wythe et al., 2013). Three further peptides that interact exclusively with the inflamed vasculature were identified in an independent phage screen. Two of these (CLDNQRPKC and CDCRGDCFC) directly inhibited angiogenesis and ameliorated AA in a Lewis rat model. The other (CRNADKFPC) was relatively inert (Yang et al., 2011) but could feasibly be attached to a therapeutic payload to facilitate delivery to the synovium, as seen with syETP‐IL‐4.

Synovial neovasculature targeting has also been achieved using scFv molecules. The scFv A7 bound to pericytes in both OA and RA synovial microvasculature, and successfully targeted human synovium transplanted in a SCID mouse. A7 did not bind to an array of control tissues, including inflamed tissues from Crohn's disease and psoriasis patients, indicating specificity for arthritic synovium (Kamperidis et al., 2011). Further scFv candidates L19 and G11 were shown to localize to collagen‐induced arthritis (CIA) joints. L19 was shown to interact with the alternatively spliced extra domain‐B of fibronectin (Fn) – an angiogenesis‐associated ECM marker. By fusing the anti‐inflammatory cytokine IL‐10 to L19, models of arthritis were successfully treated, whereas free IL‐10 was less effective. Previously, IL‐10 had failed to be efficacious in clinical trials owing to poor disease site targeting and retention (Keystone et al., 1998), so L19 could facilitate the translation of IL‐10 to the clinic for RA (Trachsel et al., 2007) (Table 2).

Table 2.

Antibody fragments and their identified binding partners

scFv	Binding partner	Described applications	References
A7	Human synovium/pericytes	Localization to synovial grafts in SCID mouse	(Kamperidis et al., 2011)
L19	Fibronectin ED‐B	Delivery of IL‐10 in arthritis	(Trachsel et al., 2007)
G11	Tenascin‐C	Localization to CIA joints	(Trachsel et al., 2007)
F8	Fibronectin ED‐A	Delivery of IL‐10 in arthritis IL‐4 delivery to ameliorate CIA	(Doll et al., 2013) (Hemmerle et al., 2014a)
1‐11E	ROS‐modified type‐II collagen	IL‐10 delivery TNFR2 delivery	(Hughes et al., 2014) (Hughes et al., 2010)

Current applications are listed, but these could be expanded by coupling with new payloads and encapsulation methods to broaden their applications. ED‐A, extra domain‐A; ED‐B, extra domain‐B.

F8, another scFv targeting a different Fn splice variant, extra domain‐A, has also been employed to deliver anti‐arthritic factors to angiogenic vessels. F8‐IL‐10 (Dekavil) in combination with a murine analogue of the established TNF inhibitor etanercept was shown to be more efficacious in treating CIA than either agent alone (Doll et al., 2013). Dekavil has also been purposed to reduce organ rejection during surgery by targeting surgical sites during heart allograft procedures (Franz et al., 2015). F8‐IL‐4 (Tetravil) in combination with dexamethasone was shown to successfully treat murine CIA (Hemmerle et al., 2014a). Tetravil has also been shown to target inflamed psoriatic sites, ameliorating inflammation and reducing pro‐inflammatory cytokine signatures in disease models (Hemmerle et al., 2014b). These examples demonstrate how the sub‐endothelial ECM can be utilized to deliver PDCs and ADCs specifically to sites of RA, with new applications for other diseases also being developed.

Targeting arthritis‐driven cartilage modifications

The presence of ROS in the joint is a common manifestation of arthritis, and ROS‐modified COL2A1 is found there. Autoantibodies against COL2A1 are commonly found in RA patients (Strollo et al., 2013). The presence of ROS can also be manipulated for therapeutic delivery to the diseased cartilage. Poly(propylene sulphide) (PPS) undergoes a hydrophobic to hydrophilic conversion upon encountering ROS, and the release of anti‐inflammatory curcumin from PPS capsules correlated with the dose of oxidizing agent. Furthermore, PPS capsules were effective at ameliorating peripheral arterial disease in a model of hind limb ischaemia in diabetic mice and were retained locally (Poole et al., 2015). Theoretically, PPS capsules could also be loaded with anti‐inflammatory cargoes and targeted to ROS‐rich arthritic joints, where the drug would be released. Furthermore, functionalization of PPS capsules by incorporating the COL2A1‐binding peptide WYRGRL permitted their retention in the cartilage ECM for much longer than the scrambled peptide, YRLGRW (Rothenfluh et al., 2008). This may further improve the therapeutic indices of drugs, by loading them into COL2A1/ROS dual‐targeting PPS capsules.

ROS‐modified COL2A1 can also be directly targeted for RA joint‐specific therapeutic delivery using scFv technology. Viral IL‐10 and soluble murine etanercept (mTNFR2‐Fc) have been successfully fused to the antibody fragment 1‐11E, which has tropism for ROS‐modified arthritic cartilage. Fusions of vIL‐10 or mTNFR2‐Fc to 1‐11E were both efficacious in animal models of arthritis (Hughes et al., 2010; Hughes et al., 2014). Feasibly, the characterization of other candidate autoantibodies against other ECM neoepitopes could permit the production of novel scFv molecules with joint tropisms for therapeutic delivery purposes. In collagen‐induced arthritis (CIA), for example, aggrecan catabolism by aggrecanase (AGG) yields the cleavage product NITEGE, which precedes the appearance of the MMP‐driven VDIPEN neoepitope (van Meurs et al., 1999). Antibodies interacting with these epitopes may allow the delivery of therapeutics to different stages of disease. Autoantibodies have also been found against other citrullinated ECM molecules of the joint including tenascin‐C (TenC) and Fn (van Beers et al., 2012; Schwenzer et al., 2016). Their characterization and conjugation to therapeutic agents could further increase the treatment arsenal for ECM delivery in arthritis.

Targeting catabolic ECM processes with enzymatically activated prodrugs

The ECM is in a constant state of turnover, with carefully orchestrated anabolic and catabolic processes occurring concomitantly. A shift towards hyperactive catabolism of the articular cartilage is a hallmark of RA, with fibroblast‐like synoviocytes, chondrocytes and myeloid cells all contributing to pathogenesis (Smolen and Steiner, 2003). MMPs are key contributors to joint destruction in both RA and OA (Yoshihara et al., 2000). There are issues with direct MMP inhibition to treat arthritis due to enzyme redundancy in cartilage destruction and reported issues with patient safety (Levitt et al., 2001). However, a strategy utilizing MMPs to release therapeutic agents at disease sites is very powerful in effect, using disease severity against itself. An early example of an MMP‐activated prodrug purposed the anthrax toxin to kill cancer cells. Anthrax toxin comprises three components: protective antigen (PA), lethal factor (LF) and oedema factor (EF). PA requires cleavage by furin‐like proteases on the cell surface before LF and EF can be internalized and elicit cytotoxicity. Substitution of the furin cleavage site of PA to one digestible by MMP‐2 (PLGMLS) or MMP‐9 (PLGLWA) yielded an MMP‐activated anthrax toxin with tumour cell specific cytotoxicity (Liu et al., 2000).

Enzymatically activated pro‐cytokines for arthritis

MMP‐activated prodrugs are also being developed to treat arthritis. The anti‐inflammatory and developmental regulator TGFβ has three mammalian isoforms (TGFβ1‐TGFβ3) exhibiting high homology in their active regions (70–80%), and each has a prodomain known as its latency‐associated peptide (LAP). Following intracellular cleavage by furin, LAP and the active moiety non‐covalently associate to form the small latent complex. Novel pro‐cytokines have been engineered by fusing the LAP of TGFβ1 to cytokines via an MMP cleavable sequence (Figure 3A), replacing the native furin site (Adams et al., 2003). LAP sterically hinders interactions between the cytokine and its receptors until removal by MMPs at sites of ECM remodelling, such as arthritic joints.

Several critical features of LAP‐fusion proteins aid their ECM targeting. Cysteines‐224 and‐226 both orchestrate dimerization of the LAP prodomain, which is critical for conferring latency (Brunner et al., 1989; Adams et al., 2003). Additionally, mutation of cysteine‐33 to serine attenuates interactions of LAP‐fusion proteins with latent‐TGFβ binding protein, preventing formation of the ECM‐deposited large latent complex and allowing their entry into the circulation (Adams et al., 2003). Finally, LAP also contains an RGD domain shown to interact with several integrins, including αVβ6 (Shi et al., 2011).

Cytokines fused to LAP are favourable in several ways for treating arthritis. First, side effects are reduced since the payload is latent until released at the disease site. Additionally, like it does for TGFβ1 (Wakefield et al., 1990), LAP increases the in vivo half‐life of its cargo, allowing longer therapeutic effects and less frequent dosing. As proof‐of‐concept, the anti‐arthritic cytokine IFNβ (Triantaphyllopoulos et al., 1999) was applied to the system. Fusions of LAP to IFNβ via the MMP‐cleavage site PLGLWA (LAP‐mmp‐IFNβ) required processing by MMP‐1, MMP‐3 or synovial fluid from RA or OA patients before in vitro activity was observed. LAP substantially increased the in vivo half‐life of IFNβ, and the molecule also ameliorated CIA (Adams et al., 2003). AGG is implicated in early, irreversible cartilage damage (van Meurs et al., 1999) and is linked more exclusively to arthritic disease than MMPs, which are also seen in cancers (Liu et al., 2000). Linking LAP to IFNβ via the AGG cleavable site DVQEFRGVTAVIR (LAP‐agg‐IFNβ) allowed slower, sustained release of the IFNβ cargo. This improved the amelioration of CIA compared to LAP‐mmp‐IFNβ (Mullen et al., 2014a). Furthermore, using an Mx2‐luciferase reporter mouse (where luciferase is expressed in response to IFN), activity of LAP‐agg‐IFNβ following intraperitoneal delivery was only observed in inflamed paws, with negligible activity in non‐inflamed paws, the liver or the serum (Mullen et al., 2014a). This strongly suggests that delivery of cytokines when fused to LAP could reduce side effects, and therapeutic targeting to disease sites can be readily achieved. This may permit the re‐evaluation of therapeutics, which failed clinical trials either due to low efficacy or high side effects.

Targeting regenerative factors to heal cartilage and bone lesions could be revolutionary in OA and RA management. To this end, the LAP platform has also been used for the encapsulation of chondrogenic GFs. LAP‐mmp‐IGF‐1 has been produced and demonstrated latency and MMP‐mediated activation. Additionally, BMP‐7 was fused to LAP, released by MMPs in vitro and elicited chondrogenic activity in a chicken mesenchymal stem cell (MSC) micromass model (Mullen et al., 2014b). This could provide a novel approach for targeted cartilage regeneration in OA and RA, stimulating resident MSCs and chondrocytes to repair the damaged ECM. Additionally, LAP has been fused to other anti‐inflammatory proteins. LAP‐fusions of the cytokines IL‐4, IL‐10 and IL‐1ra have been produced, which successfully conferred latency to them until activation by MMPs (Mullen et al., 2014b). The anti‐inflammatory peptides vasoactive intestinal peptide (VIP), α‐melanocyte‐stimulating hormone (αMSH) and γMSH have also been produced as LAP‐fusion proteins and reduced leucocyte recruitment in a peritonitis model. LAP‐mmp‐VIP, LAP‐mmp‐αMSH and LAP‐mmp‐γMSH also improved the histological state and reduced pro‐inflammatory cytokine production in a CIA mouse model (Vessillier et al., 2012). VIP causes tachycardia at high doses meaning that delivering the correct dose is critical; too low and there will be no therapeutic effect, yet too high and there is a high risk of serious adverse events. Both problems may be attenuated through fusing VIP to LAP.

Enzymatically activated intracellular delivery using protein transduction domains

MMP‐mediated intracellular delivery can also be achieved with the LAP technology. The NFκB inhibitor NEMO‐binding domain (NBD) was fused to LAP via the protein transduction domain (PTD) TAT3. Many PTDs can interact with cell surface HSPGs, which aids their uptake, along with fused payloads. The construct LAP‐mmp‐TAT3‐NBD was cleaved by MMP‐2, inhibited NFκB signalling and was effective in the treatment of carrageenan‐induced paw inflammation and antigen‐induced inflammation models (Koutsokeras et al., 2014). In this way, it may be possible to deliver anti‐inflammatory peptides intracellularly in MMP‐rich environments like the RA joint in order to reduce cytokine and degradative enzyme production.

Enzymatically activated pro‐antibodies

Antibodies to block soluble mediators show efficacy in arthritic disease, with antibodies against TNFα in particular revolutionizing RA management (Elliott et al., 1994). However, since TNFα is needed for the immune system to function, its systemic blockade has presented an increased risk of infections, as well as reactivations of latent tuberculosis (Murdaca et al., 2015). Additionally, it has been postulated that the lack of efficacy of anti‐TNFα in some patients arises from limited bioavailability at the disease site. The activatable dual‐variable domain antibody format (Wu et al., 2007) could address these issues. By fusing an anti‐intercellular adhesion molecule (ICAM)‐1 Fv region to the anti‐TNFα antibodies adalimumab or infliximab, it was possible to target the vasculature without eliciting TNFα blockade. Subsequently, by fusing the anti‐ICAM‐1 and anti‐TNFα Fv regions together using the MMP cleavable linker PLGLWA (Figure 3B), restoration of TNFα engagement was achieved following MMP cleavage (Onuoha et al., 2015). Clinically, this permits the targeting of anti‐TNFα agents to the arthritic vasculature where MMPs locally switch antibody tropism towards a therapeutic anti‐TNFα function. Using a similar concept, LAP was fused to the antigen‐binding Fv domain of antibodies via an MMP‐2 cleavable linker, and both anti‐TNFα or anti‐EGFR antibodies elicited reduced ligand engagement until cleavage and attenuation of LAP‐mediated steric hindrance (Chen et al., 2017). These LAP‐fused pro‐antibodies were stable in serum, suggesting that it may be possible to administer them systemically and achieve local delivery (Chen et al., 2017). Such modifications could improve the therapeutic index and safety of anti‐TNFα agents.

Future perspectives – what else can rheumatology learn from cancer?

Since targeted therapies for cancer are more richly researched, can we adopt any concepts from oncology to apply to arthritis treatment? The tumour‐associated ECM marker TenC is also becoming increasingly associated with RA (Schwenzer et al., 2016). The antibody F16 has been employed to target domain A1 of TenC and was efficacious at treating glioblastoma and carcinomas when conjugated to the cytotoxic agent monomethyl auristatin E via a labile valine‐citrulline dipeptide (Gébleux et al., 2017). TenC is also associated with adenocarcinomas, squamous cell carcinomas of the lung and cancer‐associated fibroblasts (CAFs) in the stroma. The TenC‐binding peptide FHKHKSPALSPVGGG localized to tumour cell xenografts in vivo and bound to biopsies from lung cancer patients (Kim et al., 2012). Incorporation of this peptide into navitoclax‐loaded liposomes successfully targeted tumour xenografts in nude mice, eliminating CAFs (Chen et al., 2016). TenC‐targeted anti‐inflammatories or GFs may have a place in RA treatment, either by directly linking therapeutic agents to targeting peptides/antibodies or conjugating a drug‐loaded capsule to a TenC targeting moiety. Moreover, the TenC binding scFv F16 also binds to atherosclerotic plaques, suggesting that it could deliver therapeutic agents to sites of atherosclerosis (Pedretti et al., 2010).

There are also some MMP‐mediated delivery technologies from oncological research, which may be repurposed for delivering drugs or nucleic acids to aberrantly behaving cells in RA. Similarly to how TAT3 was employed to deliver NBD (Koutsokeras et al., 2014), PTDs can be conjugated to drug molecules to facilitate intracellular delivery. To avoid entry into undesired cells, PTDs have been fused to polyanionic peptides via an MMP‐hydrolysable site (PLGLAG), which only enter cells following cleavage and release of the polyanion. Tumour cell‐specific delivery of the chemotherapeutic agent doxorubicin was achieved in this way (Shi et al., 2012). A method of specific MMP‐mediated nucleic acid delivery has also been developed. Nanocapsules incorporating a polycationic siRNA condensing core motif encapsulated by an MMP‐7 hydrolysable PEGylated peptide VPLSLYSGCG (Figure 3C) could deliver siRNA to the MMP‐7‐expressing cancer cell line, MDA‐MB‐231 (Li et al., 2013). By incorporating arthritis‐specific cleavable sites and appropriate anti‐inflammatory agents into these systems, they may be employed to deliver drugs or siRNA/DNA to cells in the arthritic joint and modulate their aberrant behaviours.

Concluding remarks

The ECM is being increasingly used as a target for the treatment of arthritis through direct interactions with cartilage and synovium, or drug activation by catabolic disease processes. In RA, technologies to deliver drugs to the sub‐endothelial ECM of angiogenic vessels may prove to be advantageous over previous treatment approaches. In OA, the cartilage has proven to be a difficult tissue to deliver therapeutics to due to the poor local blood supply and cell responsible for tissue homeostasis, the chondrocyte, being buried deep in the tissue. There are now several cartilage targeting peptides, which could be directly conjugated to drugs or drug‐loaded capsules to facilitate their retention within diseased cartilage. The ADCs, PDCs and materials described here will aid tissue retention of the therapy.

We may indeed be some way off able to systemically administer targeted therapeutics since there have been relatively poor deliveries achieved, and only a few such compounds are licenced for clinical use. However, there are examples of how drug therapeutic indices can be improved by targeting their delivery. The studies described certainly show progress in terms of increased efficacy and tissue retention, and so longer remission periods are possible with less frequent drug administrations. For its ultimate success, however, it is critical that the promising technologies from in vitro and preclinical studies can be translated into the clinic.

Nomenclature of targets and ligands

Key protein targets and ligands in this article are hyperlinked to corresponding entries in http://www.guidetopharmacology.org, the common portal for data from the IUPHAR/BPS Guide to PHARMACOLOGY (Harding et al., 2018), and are permanently archived in the Concise Guide to PHARMACOLOGY 2017/18 (Alexander et al., 2017c, 2017a, 2017b).

Conflict of interest

The authors declare no conflicts of interest.

Schultz, C. (2019) Targeting the extracellular matrix for delivery of bioactive molecules to sites of arthritis. British Journal of Pharmacology, 176: 26–37. 10.1111/bph.14516.