Gene Therapy for Osteoarthritis: Pharmacokinetics of Intra-Articular Self-Complementary Adeno-Associated Virus Interleukin-1 Receptor Antagonist Delivery in an Equine Model

Abstract

Toward the treatment of osteoarthritis (OA), the authors have been investigating self-complementary adeno-associated virus (scAAV) for intra-articular delivery of therapeutic gene products. As OA frequently affects weight-bearing joints, pharmacokinetic studies of scAAV gene delivery were performed in the joints of the equine forelimb to identify parameters relevant to clinical translation in humans. Using interleukin-1 receptor antagonist (IL-1Ra) as a secreted therapeutic reporter, scAAV vector plasmids containing codon-optimized cDNA for equine IL-1Ra (eqIL-1Ra) were generated, which produced eqIL-1Ra at levels 30- to 50-fold higher than the native sequence. The most efficient cDNA was packaged in AAV2.5 capsid, and following characterization in vitro, the virus was injected into the carpal and metacarpophalangeal joints of horses over a 100-fold dose range. A putative ceiling dose of 5 × 10¹² viral genomes was identified that elevated the steady-state eqIL-1Ra in the synovial fluids of injected joints by >40-fold over endogenous levels and was sustained for at least 6 months. No adverse effects were seen, and eqIL-1Ra in serum and urine remained at background levels throughout. Using the 5 × 10¹² viral genome dose of scAAV, and green fluorescent protein as a cytologic marker, the local and systemic distribution of vector and transduced cells following intra-articular injection scAAV.GFP were compared in healthy equine joints and in those with late-stage, naturally occurring OA. In both cases, 99.7% of the vector remained within the injected joint. Strikingly, the pathologies characteristic of OA (synovitis, osteophyte formation, and cartilage erosion) were associated with a substantial increase in transgenic expression relative to tissues in healthy joints. This was most notable in regions of articular cartilage with visible damage, where foci of brilliantly fluorescent chondrocytes were observed. Overall, these data suggest that AAV-mediated gene transfer can provide relatively safe, sustained protein drug delivery to joints of human proportions.

Introduction

Osteoarthritis (OA) is a chronic, erosive, often debilitating condition common to large weight-bearing joints. An estimated 30–40 million people in the United States have symptomatic OA. Yet, despite its prevalence, there are no pharmacologic agents capable of halting the underlying disease processes.¹ Although analgesics can alleviate joint pain, they cannot block the degenerative progression of OA, and many patients progress to end-stage disease, with joint replacement surgery as the only medical option.

Over the last decade, the understanding of OA pathogenesis at the cellular and molecular levels has advanced considerably,² and several naturally occurring proteins with potential as disease-modifying agents have been identified. Among the more promising candidates is interleukin-1 receptor antagonist (IL-1Ra), which functions as a competitive inhibitor of interleukin-1 (IL-1),³ a primary inflammatory cytokine and mediator of cartilage degradation.^4,5 Unfortunately, the administration of the recombinant forms of proteins, such as IL-1Ra (Anakinra), has not met with clinical success in OA due to difficulties achieving and sustaining effective levels in joints with chronic disease.⁶

Gene-based treatment approaches may provide an effective alternative.^7–9 By transferring the cDNAs encoding anti-arthritic proteins to cells resident in the tissues of diseased joints, and providing for high levels of independent expression, the biosynthetic machinery of the modified cells can be directed to overproduce and secrete the transgenic protein into the synovial fluid and surrounding tissue continuously.¹⁰ This provides the greatest concentration of therapeutic protein specifically at the site of disease, and eliminates the need for repeated application.⁹

Many vector systems have been evaluated for intra-articular gene transfer, but those based on adeno-associated virus (AAV) currently appear the most viable for clinical studies in OA. The benefits of this system include the low immunogenicity of transduced cell populations, the capacity for sustained transgene expression in the joints of laboratory animals,^11,12 and a track record of safety in clinical trials for other diseases.^13–15 Advances in AAV technology, including self-complementary (sc) vectors^16,17 and improved methods of vector production, further enhance its potential for mainstream clinical use in joint disease.

Building from positive results with local scAAV.IL-1Ra delivery in rabbits with experimental arthritis,¹⁸ the authors have worked to investigate AAV gene transfer on a scale relevant to application in large human joints. In this regard, the horse offers several advantages as a model system. The intercarpal and metacarpophalangeal (MCP) joints of the equine forelimb are similar in size and function to the human knee, and as the forelimbs carry the majority of the horse's weight during locomotion, these joints are also highly vulnerable to OA secondary to trauma and excessive loading.¹⁹ Moreover, the diagnostic technologies used to assess joint pathology in OA are the same for both humans and horses.

In pilot studies using green fluorescent protein (GFP) and human (h)IL-1Ra as reporter genes, several AAV capsid serotypes efficiently transduced fibroblasts in the synovial lining of the equine joint, and to a lesser extent chondrocytes in articular cartilage, were found.²⁰ Delivery of the human cDNA generated significant levels of hIL-1Ra in the equine synovial fluids, but following a peak at 1–2 weeks post injection, hIL-1Ra expression diminished and, after 7 weeks, was essentially undetectable by enzyme-linked immunosorbent assay (ELISA). This was attributed to immune recognition of the transduced cells expressing the xenogeneic (human) protein.²⁰

Using the equine model, this study worked to generate a detailed pharmacokinetic profile of homologous gene delivery of scAAV.IL-1Ra gene delivery in a large mammalian joint. Issues directly relevant to clinical translation, such as dosing, the levels and persistence of IL-1Ra expression, and the effect of the OA environment on intra-articular gene transfer and vector biodistribution, are addressed.

Results and Discussion

Objectives and study design

Using a codon-optimized cDNA for the equine (eq) IL-1Ra orthologue as a homologous transgene and scAAV as a delivery vehicle, over a 100-fold range from 5 × 10¹⁰ to 5 × 10¹² vector genomes (vg), first a series of dosing studies targeting the intercarpal and MCP joints of healthy horses were performed. eqIL-1Ra levels in synovial fluid, blood, and urine were measured periodically over a period of 6 months and compared to pre-injection levels. Then, to determine the impact of the pathologies characteristic of OA joints on intra-articular gene transfer, using scAAV.GFP as a cytologic marker, the local and systemic distribution of the vector and transduced cell populations following vector delivery into healthy joints and those with late-stage, naturally occurring OA were examined.

Summary of data

scAAV.eqIL-1Ra vector construction and characterization in vitro and in vivo

In a study performed >15 years ago, measurable levels of transgenic eqIL-1Ra production in equine joints were reported following intra-articular injection of a first-generation recombinant adenoviral vector.²¹ More recently, as validated equine-specific ELISA reagents became available, protein expression from the native cDNA (NCBI: NM_001082525.2) was found to be somewhat modest. Thus, to minimize immune recognition of the IL-1Ra transgene product²² and maximize expression, the eqIL-1Ra cDNA codon was optimized²³ and synthesized both with and without a consensus Kozak sequence²⁴ leader immediately upstream of the translation start site. Following insertion of the cDNAs into the scAAV vector plasmid, pHpa-trs-SK,^16,17 in transient transfection assays, both optimized constructs produced eqIL-1Ra at levels 30- to 50-fold higher than the native sequence (Fig. 1a). As the construct with the Kozac sequence consistently provided the highest eqIL-1Ra expression in these assays, it was selected for viral packaging and testing in vivo.

Figure 1.

Expression and function of the codon-optimized eqIL-1Ra transgene in vitro. (a) eqIL-1Ra in conditioned media at 48 h following transfection of equine synovial fibroblasts with the Hpa-trs-sk, scAAV vector plasmid containing the coding sequences for GFP, native eqIL-1Ra, codon-optimized eqIL-1Ra (Opt), and codon-optimized eqIL-1Ra with a Kozak sequence leader (K + Opt). n = 3 transfections. (b) eqIL-1Ra in conditioned media following infection of equine synovial fibroblasts with increasing doses of scAAV.eqIL-1Ra packaged in serotype 2.5. Viral genomes (vg) per cell of scAAV.eqIL-1Ra are shown to the right of the respective expression curves. Parallel infection with 10⁵ vg/cell of scAAV.GFP was used as a negative control. n = 3 infections. (c) PGE₂ production following IL-1 stimulation of cultured equine synovial fibroblasts that were uninfected (no virus), or transduced with 10⁵ vg/cell of scAAV.GFP or scAAV.eqIL-1Ra 48 h previously. Following the addition of IL-1 to the growth media to final concentrations of 0, 1, 5, or 10 ng/mL, the conditioned media were collected and assayed for PGE₂ by ELISA. Each bar represents the mean of n = 3. Error bars represent ±SEM. eqIL-1RA, equine interleukin-1 receptor antagonist; scAAV, self-complementary adeno-associated virus; GFP, green fluorescent protein; PGE₂, prostaglandin E₂; IL-1, interleukin-1; ELISA, enzyme-linked immunosorbent assay; SEM, standard error of the mean.

The previous data showed human synovial fibroblasts in culture to have a preference for infection with AAV2.²⁰ Considering possible translation to human testing, the eqIL-1Ra vector construct was packaged in the AAV2.5 capsid,²⁵ which maintains AAV2 tropism but shows reduced reactivity with AAV2 neutralizing antibody, prevalent among the human population.²⁶ Infection of equine synovial fibroblast cultures with a range of doses of the AAV2.5 vector resulted in exceptionally high expression of eqIL-1Ra, which exceeded 10 μg/mL at 10⁵ DNAse resistant viral genomes (vg)/cell (Fig. 1b). eqIL-1Ra production did not exceed background in parallel control cultures infected at 10⁵ vg/cell with an scAAV2.5 vector containing GFP (scAAV.GFP). The function of the transgenic eqIL-1Ra protein was confirmed in bioassays, where around a 90% reduction in prostaglandin E₂ (PGE₂) levels was seen in media conditioned by equine synovial cells previously infected with AAV.eqIL-1Ra at 10⁵ vg/cell and subsequently challenged with increasing amounts of recombinant human IL-1 (Fig. 1c).

To determine the effect of vector dose on eqIL-1Ra expression following intra-articular delivery, an approach was designed to provide insight into the intra- and inter-animal variability while using a minimum number of experimental animals. In each of six horses, injections of scAAV.eqIL-1Ra were distributed in random order among both intercarpal and MCP joints of both forelimbs at three different doses: 5 × 10¹⁰, 5 × 10¹¹, and 5 × 10¹² vg; Fig. 2a). The remaining joint was injected with an equivalent volume of the saline delivery vehicle (Lactated Ringer's solution) and served as a negative control. Immediately prior to vector injection, peripheral blood was drawn, urine was collected, and synovial fluid was aspirated from each of the four joints. Periodically thereafter, over a predetermined interval of 6 months, synovial fluid, blood, and urine were similarly collected from each animal. The eqIL-1Ra content in the recovered fluids was measured using commercially available ELISA, and was assessed relative to pre-injection values.

Figure 2.

scAAV administration and expression in the equine joint. (a) Right: Anatomic locations of the carpal (Ca) and metacarpophalangeal (MCP) joints of the equine forelimbs (black arrows). Left: Radiographic image of the equine carpus, showing the locations of the intercarpal (white arrow), carpometacarpal (circle), and antebrachiocarpal (asterisk) joints. The synovial cavities of the intercarpal and carpometacarpal joints communicate. The locations of the radiocarpal (RC) and third carpal (C3) bones discussed in Fig. 3a and b are shown as indicated. (b) eqIL-1Ra in synovial fluids of forelimb joints injected with either saline vehicle (Lactated Ringer's solution) or vehicle containing scAAV.eqIL-1Ra at vg doses shown to the right of the respective plots (n = 6 joints). Error bars represent ±SEM. The mean IL-1Ra levels at each dose and time point were analyzed using independent sample t-tests. (a) IL-1Ra expression at 5 × 10¹² and 5 × 10¹¹ vg doses is greater than 5 × 10¹⁰ vg dose; p < 0.05. (b) IL-1Ra expression at 5 × 10¹² vg dose is greater than 5 × 10¹¹ vg dose; p < 0.05.

Despite receiving AAV in three forelimb joints, no adverse effects were observed acutely or at any point during the protocol, and eqIL-1Ra levels in both serum and urine remained at baseline levels. Among the control joints injected with the fluid vehicle, synovial fluid eqIL-1Ra also remained at pre-injection levels (<1 ng/mL) throughout (Fig. 2b). In joints receiving the AAV vector, dose-related increases in synovial fluid eqIL-1Ra were observed within 2 weeks of injection, with mean levels ranging from around 6 ng/mL at 5 × 10¹⁰ vg to around 40 ng/mL at 5 × 10¹² vg. Peak eqIL-1Ra production occurred between 4 and 8 weeks post injection, and these levels were maintained for the remainder of the 6-month study. In contrast to the near-linear relationship between vector dose and eqIL-1Ra expression seen in culture (Fig. 1b), over the dose range tested in vivo, a 100-fold increase in vector was required to produce around a sevenfold increase in synovial fluid eqIL-1Ra. Moreover, at the high end of the dose range, the 10-fold increase in vector dose from 5 × 10¹¹ to 5 × 10¹² vg, only increased the mean eqIL-1Ra level by around 50%, which was not statistically significant (Fig. 2b). These data suggest a possible ceiling effect²⁷ such that additional increases in vector dose would be unlikely to generate a meaningful increase in synovial fluid eqIL-1Ra, at least within the context of a healthy joint.

AAV2.5 transgene expression in healthy versus OA joints

As the 5 × 10¹² vg dose of scAAV.eqIL-1Ra consistently provided the highest eqIL-1Ra production intra-articularly, seemed reasonably safe, and appeared to represent a maximum functional dose, it was selected for further characterization in vivo. To gain insight into the safety of intra-articular AAV gene delivery and thus its potential for clinical application in OA, the scAAV.GFP vector was first packaged in the AAV2.5 capsid. Then 5 × 10¹² vg of scAAV.GFP was injected into one intercarpal joint of three healthy horses and three horses with advanced naturally occurring OA (Fig. 3a and b). Two weeks post injection, the horses were euthanized, and tissues from the injected joints and nine additional sites throughout the body (e.g., the brain, heart, liver, and spleen) were collected for analysis. GFP fluorescence was used as a cytologic marker to determine the influence of the characteristic OA pathologies (Fig. 3b) on the local and systemic distribution of the of the transduced cell populations relative to healthy joints. Quantitative polymerase chain reaction (qPCR) was used to quantify the scAAV genome copies in the various tissues to determine the vector biodistribution.

Figure 3.

GFP expression in equine synovium from healthy and osteoarthritis (OA) joints following intra-articular gene delivery with scAAV. One intercarpal joint of three healthy horses and three with advanced, naturally occurring OA was injected with 5 × 10¹² vg of scAAV.GFP. Two weeks later, the joint tissues were collected and analyzed for fluorescence. (a and b) Opposing articulating surfaces from representative healthy (a) and OA (b) intercarpal joints are shown in the same orientation. The locations of the radiocarpal (RC) and third carpal (C3) bones are as indicated in (a). Full-thickness cartilage erosions on the surfaces of both the RC and C3 bones are indicated by black arrows; osteophytes along the margins of the C3 bone are indicated with asterisks. Bone and cartilage damage from an osteochondral fracture is indicated by the red arrow. (c and d) scAAV.GFP expression in the synovium of a healthy joint. (e) Hematoxylin and eosin (H&E) stain of normal synovial membrane shows the surface (intimal) layer composed of single row of flat and nonreactive cells overlaying vascularized fibro-adipose tissue. (f and g) GFP expression across the synovial surface of an OA joint. (c and f) Direct fluorescence in freshly harvested tissues at 10 × . (d and g) Paraffin sections immunohistochemically stained for GFP at 20 × .) (h) H&E stain of synovial membrane from an OA joint shows an irregular intimal lining and markedly increased cellularity due to hypertrophy and hyperplasia of the fibroblasts and a chronic inflammatory infiltrate (primarily lymphocytes). (i) Direct fluorescence microscopy of synovium from a joint that was not injected with scAAV.GFP and serves as a control for autofluorescence. (j) Control synovium following immunohistochemical stain for GFP expression reflects the specificity of the antibody and the absence of background staining. (k) Representative GFP expression (arrows) in fresh synovial tissue harvested from the antebrachiocarpal joint of the carpus following injection of scAAV.GFP into the intercarpal joint. These compartments do not communicate within the carpal joint. (The images shown here and in Fig. 4 are composites from the six animals in both groups. As the fresh-tissue samples were of variable thickness and composition, a wide range of exposure times was used to capture the varying fluorescence intensities at the different magnifications indicated. The contrast and brightness of individual panels were adjusted linearly for uniformity of appearance and to reflect fluorescence intensities viewed by direct microscopy.)

Similar to results with AAV serotypes tested previously,²⁰ in each of the healthy joints, the predominant site of GFP expression was the synovium (Fig. 3c–e). Examination of the freshly harvested tissues revealed abundant fluorescent cells throughout the lining of the joint capsules of the intercarpal and the carpometacarpal joints, which was expected, as the synovial cavities of these joints communicate (see Fig. 2a). In both joints, the majority of the fluorescently labeled cells appeared to be concentrated in the thicker villous regions (Fig. 3c). When viewed in cross-section under increased magnification, the modified cells were comprised primarily of non-reactive fibroblasts residing in a thin synovial membrane (Fig. 3d and e). GFP activity was also visible in articular cartilage shavings from both joints, but was generally faint and limited to scattered isolated cells typical of that seen in Fig. 4a and b. In striking contrast, GFP activity in the synovium of the OA joints was much higher than that seen in healthy joints. These samples were often brilliantly fluorescent, even at low magnification (Fig. 3f). While the density of the fluorescent cells was visibly higher across the entire expanse of the lining of the OA joints, it was particularly so in regions with marked fibroblast hyperplasia and leukocytic infiltration typical of chronic inflammation (Fig. 3g and h). Similar to healthy joints, the fluorescent cells were almost exclusively delimited to the synovium and subsynovium, and only rarely seen in the supporting fibrous tissues.

Figure 4.

GFP expression in equine cartilage from healthy and OA joints following intra-articular gene delivery with scAAV. The intercarpal joints of three healthy horses and three with late-stage, naturally occurring OA were injected with 5 × 10¹² vg of scAAV.GFP. Two weeks later, the joint tissues were collected and analyzed for fluorescence. (a and b) GFP expression in fresh cartilage shavings immediately following harvest from a healthy joint. Arrows indicate representative fluorescent cells. (c) GFP activity in healthy cartilage shaving after 48 h in explant culture. (d–f) GFP activity in fresh cartilage shavings from an OA joint. (e and f) Magnified fields from (d) (d: 10 × ; e and f: 20 × ). (g and h) GFP activity in fresh OA cartilage from regions with visible erosion (g: 10 × ; h: 20 × ). (i) GFP+ chondrocytes changing morphology in degrading cartilage matrix (paraffin section; 40 × ). (j) GFP expression in chondrocyte clusters in fresh cartilage shavings OA joint (40 × ). (k) GFP expression in cartilage clusters from an OA joint (paraffin section). (l) H&E-stained section of OA cartilage similar to (k). (m) Cartilage section from a naïve joint immunostained for GFP (negative control for positive staining in i and k). (n and o) GFP activity in freshly harvested osteophyte (n: 10 × ; o: 20 × ).

Relative to the articular tissues in the healthy joints, the OA cartilage showed the most dramatic increase in GFP activity, as populations of brightly fluorescent cells were readily apparent in all shavings recovered (Fig. 4d–f). The labeled chondrocytes included both elongated cells, consistent with superficial layer chondrocytes (Fig. 4e), and cells with more spherical morphology characteristic of chondrocytes in deeper layers (Fig. 4f). Shavings harvested near full-thickness erosions often contained focal regions with intense GFP fluorescence readily visible at low magnification (Fig. 4g and h). These areas frequently contained cells with spindle-shaped morphology, consistent with dedifferentiation to fibrochondrocytes (Fig. 4h and i).²⁸ GFP expression was particularly prominent in chondrocyte clusters characteristic of OA cartilage (Fig. 4j–l). Pockets of brightly fluorescent cells were also visible across the surfaces of osteophytes (Fig. 4n and o) recovered from the margins of the OA joints (Fig. 3b).

For both healthy and OA joints receiving virus, the synovium of the antebrachiocarpal joint of the carpus was the only tissue outside the injected joint noted to contain visible GFP fluorescence (Fig. 3k). Despite its location immediately proximal to the site of injection (Fig. 2a, *), only sparse fluorescent cells were seen and in only a few of the harvested samples.

In a related observation, it was found that the GFP expression in shavings of healthy cartilage (barely detectable at harvest; Fig. 4a and b) increased dramatically after 48 h of incubation in explant culture, such that dense populations of vividly fluorescent cells appeared throughout the matrix in each sample (Fig. 4c). This suggested that a large percentage of the chondrocytes had actually been transduced by the virus in situ, but failed to express the fluorescent reporter protein above the visible threshold in the context of the healthy joint.

AAV2.5 biodistribution

To assess the emigration of the AAV vector from the joint following intra-articular injection, total DNA was isolated from the articular and extra-articular tissue samples harvested from the animals injected with scAAV.GFP and assayed for vector genome content by qPCR. As shown in Table 1, the systemic distribution of vector genomes was largely consistent with visible GFP activity. For all animals, the tissues from the joints receiving virus showed the highest vector genome content. Though there was wide variation among individuals, on average, the vector DNA measured in the synovium was around 30- to 50-fold higher than in cartilage, with no significant differences between OA and normal joints. Detectable but considerably fewer vector genomes were found in the synovium of the adjacent antebrachiocarpal joint in horses from both groups. Outside the carpus, one animal from the healthy group showed a low number of vector genomes in the liver, while the liver and spleen from one animal in the OA group were also positive for vector content. Altogether, under both healthy and diseased conditions, >99.7% of the AAV vector genomes detected were in the joint tissues. These results indicate that the AAV vector primarily remains within the injected joint, and while the OA environment appears to substantially enhance transgene expression intra-articularly, it does not appear to impact extra-articular vector dispersion meaningfully.

Table 1.

Distribution of AAV.GFP genomes following injection in the intercarpal joint of healthy horses and those with naturally occurring osteoarthritis

	Healthy			OA
Tissue	Horse 1	Horse 2	Horse 3	Horse 4	Horse 5	Horse 6
Synovium intercarpal joint	178,313 ± 559	66,022 ± 167	16,460 ± 114	212,262 ± 237	15,590 ± 280	142,033 ± 720
Cartilage intercarpal joint	3,349 ± 43	989 ± 109	1,726 ± 59	2,499 ± 69	887 ± 19	6,177 ± 55
Antebrachiocarpal synovium	159 ± 86	20 ± 3	ND	ND	ND	84 ± 16
Peri-articular muscle	ND	ND	ND	ND	ND	ND
Contralateral intercarpal synovium	ND	ND	ND	ND	ND	ND
Contralateral quadriceps	ND	ND	ND	ND	ND	ND
Ipsilateral MCP	ND	ND	ND	ND	ND	ND
Brain	ND	ND	ND	ND	ND	ND
Heart	ND	ND	ND	ND	ND	ND
Liver	ND	ND	45 ± 4.8	380 ± 15	ND	ND
Lung	ND	ND	ND	ND	ND	ND
Spleen	ND	ND	ND	277 ± 64	ND	ND

Values represent vector genome copies per microgram of genomic DNA and are means of at least three replicates ± standard error of the mean. Tissues and locations are in reference to the injected joint.

MCP, metacarpophalangeal; ND, not detected.

Conclusions

These studies show that in a large mammalian joint, scAAV-mediated delivery of a homologous IL-1Ra cDNA is capable of elevating the steady state levels of IL-1Ra protein in the synovial fluid by >40-fold over the endogenous background. Distinct from the relatively brief expression seen previously following delivery of the xenogeneic human IL-1Ra cDNA in the equine joint,²⁰ in the absence of immune recognition and elimination, the articular cells transduced by the vector can sustain high level IL-1Ra synthesis for at least 6 months.

Favorable to the development of a gene-based therapy, the pathologic changes in tissue morphology and cellularity typical of the OA joint appear to increase viral transduction and transgenic expression substantially. Nonetheless, in both healthy and diseased environments, the vast majority of the vector DNA is retained in the injected joints. As the vector-mediated overproduction in the forelimb joints did not increase the circulating IL-1Ra levels in peripheral blood, it appears that AAV.IL-1Ra gene delivery in large-mammal joints provides little risk of immunosuppression from systemic IL-1 blockade, at least with the vector doses used here.

The expression data in healthy equine joints are largely consistent with those reported in an exploratory study by Goodrich et al.²⁹ In both cases, codon optimization of the native eqIL-1Ra sequence enhanced secretion of the encoded protein by at least 20-fold. In the Goodrich study, despite administering a similar vector dose, the transgenic IL-1Ra levels in the synovial fluid were around two- to threefold higher than observed in the current study. The basis for this is uncertain, but it may reflect the small sample size in the earlier study or possibly variations in methods for fluid collection and IL-1Ra quantification. Regardless of the discrepancies in absolute level, the overall patterns of expression are in good agreement, and independently serve to document the capacity of the AAV vector to enable sustained therapeutic transgene expression in a large synovial joint.

Local AAV-mediated gene transfer was explored previously in human joints for the treatment of rheumatoid arthritis³⁰ and entered Phase II study.³¹ Unfortunately, measurement of the transgenic protein (a tumor necrosis factor alpha antagonist) produced in the joints was not part of the protocol, and no information emerged from the study regarding the level and duration of expression that was achieved in human joints. This information, though, seems crucial to effective clinical application. To this end, the use of the equine system has proven particularly informative, allowing the capacity of an AAV vector for therapeutic gene delivery to be tested on a scale equivalent to the human knee. The ability to aspirate synovial fluid serially from the equine joint enables direct measurement of secreted transgene products from the same animal over time and the generation of reliable pharmacokinetic profiles. A particularly beneficial aspect of the horse as a model system is the ability to study gene delivery in the context of naturally occurring joint disease, rather than an experimental model.

The studies of vector biodistribution show similar results for both healthy and OA joints. The cells within the synovial lining are the primary targets for AAV gene transfer, and on average are infected around 40-fold more efficiently than chondrocytes in articular cartilage and thus are primarily responsible for the transgenic IL-1Ra content in the synovial fluid. In equine synovium, on average around 100,000 AAV vector genomes per microgram of cellular genomic DNA (gDNA), or around one vector genome per cell, was measured. This efficiency of infection is similar to the results in rats following injection in the stifle joints of weight-adjusted doses of 10⁹ and 10¹⁰ vg of scAAV.hIL-1Ra, also packaged in the AAV2.5 capsid.³² Recently, Bevaart et al.³³ described a safety study of AAV5-mediated human IFNβ gene delivery in the joints of nonhuman primates with collagen-induced arthritis. In this case, infection levels in synovium >200 vg/cell were reported, but at vector doses >150 times that used in equine joints. Considering the data in rats and horses with the sc-vector and AAV2.5 capsid, the greater tissue and fluid volume of the equine joint does not appear to affect gene delivery to synovium adversely, such that dosing and infection efficiency scale upward proportionally.

Interestingly, the IL-1Ra expression studies point to a vector ceiling dose at around 5 × 10¹² vg, and suggest that vector doses even 10-fold or more higher are unlikely to provide a significant increase in steady-state IL-1Ra levels in synovial fluid. This effect is likely a reflection of the rapid turnover of the synovial fluid and the continuous efflux of the fluid and solutes into the extra-synovial interstitium and lymphatics.^34,35 In contrast to the static conditions in culture where secreted transgene products, such as IL-1Ra, can steadily accumulate in conditioned medium, small soluble molecules have a brief residence time in synovial fluid. Thus, the increase in synovial fluid IL-1Ra content following gene delivery represent homeostatic levels in the face of ongoing dilution. Since the 5 × 10¹² vg dose was able to maintain a >40-fold increase in synovial fluid IL-1Ra content for several months, this apparent ceiling effect is unlikely to impede the efficacy of therapeutic gene delivery.

Considering therapeutic application, markedly higher expression of the GFP reporter was seen in association with the characteristic cellular and morphologic changes in the joint found in advanced OA.³⁶ For example, in the synovium, chronic inflammatory stimulation frequently induces synovial hyperplasia, angiogenesis, leukocytic infiltration, and fibrotic thickening.^37,38 While AAV-mediated GFP expression appeared consistently higher throughout the synovium of OA joints, fluorescence was especially pronounced in regions of infiltration and hyperplasia.

The most notable increase in GFP expression occurred in the OA cartilage. In stark contrast with cartilage from healthy joints where fluorescent cells were sparse and barely visible, shavings harvested from OA joints contained abundant brightly fluorescent cells throughout, with the most striking GFP activity focused at sites with obvious matrix erosion. While it would seem that physical damage to the matrix would facilitate vector diffusion into the OA cartilage, no difference in vector genome content was seen in the chondrocytes from OA and healthy joints. Instead, the data suggest that the increased transgene expression in the cartilage (and synovium as well) arises from heightened inflammatory and stress-induced activation in OA tissues.³⁹

Chondrocytes residing in healthy articular cartilage exist in a non-distressed, resting state. In OA, however, damage to the cartilage matrix diminishes its protective properties and exposes the regional chondrocyte populations to excessive mechanical loading.⁴⁰ These abnormal, pathologic forces induce stress responses in the chondrocytes, which drive the metabolism of the normally quiescent cells into a highly activated state.^41–45 The chondrocytes become proliferative, secrete high levels of inflammatory cytokines, and release proteolytic enzymes that further degrade the surrounding matrix.^44–46

AAV transduction efficiency is known to be enhanced (in some cases by several log orders) by molecular mechanisms associated with intracellular stress.^47–51 Stimuli that cause subcellular stress frequently disrupt the ubiquitin–proteasome pathway and thereby inhibit targeted degradation of AAV vector particles that accumulate in the peri-nuclear cytoplasm of infected cells. Heightened vector stability in concert with increased nucleolar import and capsid uncoating in stressed cell populations serve to amplify the number of vector genomes in the nucleus available for transgene expression.^49,51–55

In its native viral context, transcription from the cytomegalovirus (CMV) immediate early promoter is induced in response to NF-κB activation and signal transduction from p38 and other stress-activated protein kinases. Stress-related induction reactivates human CMV from latency, and is required for expression of genes necessary for DNA replication.^56–60 Inflammation and cellular stress can likewise increase transcription and expression of transgenes under control of the CMV immediate early promoter.^58,61–64 In this regard, OA cartilage is highly enriched with stress-activated chondrocytes,⁴³ especially at sites of cartilage degradation (where AAV.GFP expression was greatest). GFP fluorescence was also prominent at sites of chondrocyte proliferation and cluster formation and in osteophytes, which arise from persistent activation of chondro-osseous progenitors at the transition from cartilage to synovium.⁶⁵ Potent induction of GFP expression was also seen in cartilage shavings from healthy joints following incubation in explant culture. As no additional vector was added to the shavings, this sudden burst in GFP activity had to arise from vector DNA already present in the chondrocytes. Therefore, it must reflect a dramatic change in metabolism in infected chondrocytes from the stress of cartilage harvest and/or change in growth conditions.

A number of reports describe the generation of synthetic inflammation-inducible promoter systems for gene therapy applications.^66–68 The AAV vector discussed here, at least within the context of a large mammalian joint, appears to be highly responsive to the OA environment and innately disease activated. Moreover, the regional differences in GFP expression seen in OA cartilage indicate the potential to direct therapeutic transgene expression preferentially to cartilage regions under the greatest pathologic stress. This opens the door for development of derivative vectors for disease-targeted anabolic stimulation of cartilage repair and regeneration.

Altogether, the data from this study support consideration of scAAV.IL-1Ra for human testing. These results validate the use of AAV gene transfer for sustained protein drug delivery in large mammalian joints, and the capacity of the transduced cells to sustain transgenic protein expression at high levels for several months. Fortuitously, transgene delivery and expression appear to be considerably more efficient in the disease environment of the OA joint. As no adverse effects of treatment were seen, and the vector and IL-1Ra production appear effectively contained within the treated joints, the procedure overall seems reasonably safe, with the potential to offer benefit in joints of human proportions. The Materials and Methods section can be found in the Supplementary Data (available online at www.liebertpub.com/humc).

Acknowledgments

This project was supported by grants AR048566 and AR048566-S from the National Institute of Arthritis Musculoskeletal and Skin Diseases (NIAMS) of the National Institutes of Health. We wish to thank Lorraine Matheson and the National Gene Vector Biorepository for assistance obtaining the packaging/helper plasmids for the AAV2.5 capsid. We gratefully acknowledge the facilities that assisted with large scale preparation of the AAV vectors for the study: (1) the Vector Core of the Powell Gene Therapy Center at the University of Florida, with assistance from Mark Potter; and (2) the Vector Core of the University of North Carolina at Chapel Hill, with patient assistance from Josh Grieger. We also extend our gratitude to MaryBeth Horodyski, Darlene Bailey, and Jennifer Streshyn for their assistance and administrative support throughout the performance of the study.

The optimized equine IL-1Ra cDNA can be obtained from the corresponding author through material transfer agreement with the University of Florida. The pXR.2.5, pXX6-80, and pHpa-tr-sk (ScAAV-CMV-GFP) plasmids for packaging self-complementary vectors in the AAV2.5 capsid are available through the National Gene Vector Biorepository and material transfer agreement with the University of North Carolina at Chapel Hill.

Author Disclosure

C.H.E. and S.C.G. are inventors on several patents and patent applications describing cell- and gene-based therapies for arthritis and connective tissue disorders.