Injectable hyaluronic acid hydrogels encapsulating drug nanocrystals for long‐term treatment of inflammatory arthritis

Abstract Antiproliferative chemotherapeutic agents offer a potential effective treatment for inflammatory arthritis. However, their clinical application is limited by high systemic toxicity, low joint bioavailability as well as formulation challenges. Here, we report an intra‐articular drug delivery system combining hyaluronic acid hydrogels and drug nanocrystals to achieve localized and sustained delivery of an antiproliferative chemotherapeutic agent camptothecin for long‐term treatment of inflammatory arthritis. We synthesized a biocompatible, in situ‐forming injectable hyaluronic acid hydrogel using a naturally occurring click chemistry: cyanobenzothiazole/cysteine reaction, which is the last step reaction in synthesizing D‐luciferin in fireflies. This hydrogel was used to encapsulate camptothecin nanocrystals (size of 160–560 nm) which released free camptothecin in a sustained manner for 4 weeks. In vivo studies confirmed that the hydrogel remained in the joint over 4 weeks. By using the collagen‐induced arthritis rat model, we demonstrate that the hydrogel‐camptothecin formulation could alleviate arthritis severity as indicated by the joint size and interleukin‐1β level in the harvested joints, as well as from histological and microcomputed tomography evaluation of joints. The hydrogel‐nanocrystal formulation strategy described here offers a potential solution for intra‐articular therapy for inflammatory arthritis.

and creating pannus tissue. 6,8 Inhibiting the proliferation of synoviocytes is thus a potential treatment strategy for RA. 9 Various antiproliferative drugs, developed for cancer treatment, have been shown to be effective to treat symptoms of RA, leading to a convoluted connection between cancer chemotherapeutics and RA treatment. 10 Camptothecin (CPT), a topoisomerase I inhibitor, 11 for instance, is one such chemotherapeutic agent that can inhibit the proliferative processes in RA (e.g. synovitis and angiogenesis). 12,13 However, the in vivo application of this potent drug is limited by its extremely low solubility and instability in biological fluids and the associated high systemic toxicity. Additional obstacles for the application of CPT in inflammatory arthritis treatment, as for other small molecular drugs, are low joint bioavailability and short joint residence time. 14,15 Localized drug administration, namely intra-articular (IA) injection, has been demonstrated to be effective in improving drug's joint bioavailability and reducing systemic toxicity. 14,16 However, extra efforts have to be made to prolong the drug's residence time in the joint, given that small molecular drugs are typically cleared into systemic circulation within a few hours or less. 14 One approach toward that end is to formulate therapeutic agents into macroscopic carriers, such as polymeric microparticles, which are injectable but sufficiently large to avoid the rapid lymphatic drainage. 15 Zilretta™ (FX006), recently approved by the U.S. Food and Drug Administration (FDA), is one example that uses poly(lactic-co-glycolic acid) microspheres to encapsulate triamcinolone acetonide for IA injection. 17 While polymer particles can efficiently extend the drug release over weeks to months, the drug loading capacity is generally low (usually less than 10 wt%), 18 and as a result, a large amount of particles has to be injected into the joint to achieve therapeutic drug concentrations. [19][20][21] An alternative method to creating drug depot inside the joint is to use injectable hydrogels. Hydrogels with three-dimensional crosslinked network, high water content, and tissue-like properties offer a versatile drug delivery platform with tunable drug loading capacity and release kinetics. Hyaluronic acid (HA) hydrogels, in particular, have a track record of safe and effective clinical uses in IA therapy with various injectable HA hydrogels been approved by the FDA to provide viscosupplementation. 14 Yet, engineering hydrogels that can sustain the release of small molecular drugs in the time frames needed for treating chronic diseases such as inflammatory arthritis is still a difficult task. Encapsulating solid-form drugs, such as drug nanocrystals, instead of free drugs, can significantly extend the release period. Drug nanocrystals, as commonly used to formulate poorly water-soluble therapeutical agents, have slow dissolution rates and improved drug stability under physiological conditions. 22,23 Moreover, the compact crystalline structures are composed of nearly 100% active pharmaceutical ingredient, thus leading to a high loading efficiency. The vast potential of drug crystals in drug delivery has been gradually uncovered, with a few tens of crystalline drug formulations successfully translated to the clinic. 22 However, their applications in IA drug delivery remain largely unexplored, with only a few drug nanocrystal-based formulations reported for osteoarthritis treatment. 24,25 Here, we report a nanocrystal-encapsulated, HA hydrogel-based formulation for localized, long-term IA drug delivery (Figure 1), as such to develop a potential inflammatory arthritis treatment by using the chemotherapeutic agent CPT. To the best of our knowledge, this is the first attempt to explore IA injection of nanocrystal-loaded hydrogels as a sustained release depot for inflammatory arthritis treatment.
The first and foremost step toward this is to develop an in situforming injectable HA hydrogel that can load drug nanocrystals. While several HA products are used for IA delivery, 26,27 the prevalent crosslinking methods using butanediol-diglycidyl ether or divinyl sulfone require hostile reaction conditions (e.g. basic pH) and post-gelation washing processes, 27,28 which limit their use for in situ gelling and drug encapsulation. Here, we use a naturally occurring, highly biocompatible click reaction: cyanobenzothiazole (CBT)/cysteine reaction as the cross-linking chemistry, which is ideally suited for this purpose as it has fast kinetics and good yield under physiological conditions without the necessity of catalyst or harsh conditions. 29-31 CPT nanocrystals can be encapsulated into HA hydrogels by simply mixing them with the hydrogel precursors. The promise of CPT nanocrystal/HA hydrogel formulation in achieving long-term localized IA drug delivery was verified both in vitro and in vivo.

| Chemical synthesis and characterization of HA hydrogel precursors
An ideal cross-linking method for fabricating in situ-forming injectable HA hydrogels that can encapsulate drug nanocrystals should operate under mild physiological conditions (neutral pH in aqueous solution at 37 C) with desirable gelation kinetics and outstanding biocompatibility, such as catalyst-free, no undesired byproducts, and low toxicity of the initial, final, and degraded products. The cross-linking chemistry chosen here is the CBT/cysteine reaction, which is the last step reaction in the generation of D-luciferin in fireflies. [29][30][31] For the ease of synthesis, we conjugated CBT to HA, which improved its aqueous solubility while preserving cross-linkability when mixed with cysteinebased cross-linkers in aqueous media at a physiological pH. The primary amine of glycine(Gly)-CBT was coupled to HA carboxyl groups using EDC/sulfo-NHS chemistry (Figure 2a). Successful conjugation was verified with 1 H nuclear magnetic resonance (NMR) spectroscopy as the three characteristic CBT proton peaks at δ 7.5-8.4 ppm are shown in the final product ( Figure 2b). To further confirm the covalent attachment of the Gly-CBT to HA and the structural integrity of HA after conjugation, aqueous gel permeation chromatography (GPC) with refractive index (RI) and ultraviolet (UV) detectors was used to characterize the HA-CBT conjugates (Figure 2c). HA-CBT conjugates have a similar retention volume, that is, similar hydrodynamic volume, to blank HA (the starting material), indicating that the conjugation and purification processes do not degrade HA. This is highly desirable for its further biomedical applications as the biological functions of HA are substantially sensitive to its molecular size. 32 Moreover, in contrast to blank HA with no UV absorbance, the eluted HA-CBT products have an observable UV absorbance peak in the similar wavelength range of CBT derivatives ( Figure S1), which is a strong evidence that the CBT moieties are covalently attached to the HA backbone.
The degree of CBT conjugation (DC) was estimated by comparing the peak areas of aromatic protons from CBT (δ 7.5-8.4 ppm) and methyl groups in the acetamido moiety of HA (δ 1.7-2.0 ppm), and DC values of 5.1 ± 0.5%, 10.3 ± 0.5%, and 18.5 ± 4.4% were achieved by controlling the molar feed ratio of CBT to carboxyl groups of HA, which is highly beneficial for tuning the hydrogel mesh sizes and mechanical were designed as hydrophilic cross-linkers, which were synthesized by reacting PEG diamines (of different molecular weights) with thioland amine-protected cysteine (Figure 2d). PEG was chosen due to its biocompatibility, aqueous solubility, low viscosity, and commercial availability in different sizes. 33 The structures of the deprotected and purified PEG linkers were confirmed by matrix-assisted laser desorp-   (Figure 3a). This falls into the ideal optimal gelation time range (1-10 min) 34 for an injectable hydrogel. A continuous increase in storage modulus with time after the gel point was reached and indicates the occurrence of post-gelation cross-linking reactions. This suggests that the initially formed network percolating the entire system can be further strengthened through such continued cross-linking. This is indeed beneficial for a practical implementation of hydrogel injections, as the initial formed, loosely cross-linked hydrogels (low G 0 ) can restrict the undesired massive spread of injected polymers, but still permit its injectability (even if the gelation happened inside the syringe) and fillability for tissue cavity with arbitrary shapes. The formation of a wellstructured hydrogel network with solid-like properties can also be corroborated from the dramatic decrease of tan(δ) upon gelation and the large G 0 /G 00 ratio after gel formation ( Figure 3a). Furthermore, as shown in Figure 3b, a constant G 0 was obtained over the entire frequency range tested, and it is higher than G 00 by 1-2 orders of magnitude. The independence of G 0 on the frequency is another characteristic viscoelastic response of a cross-linked gel network.
According to the polymer network theory, G 0 is highly dependent on the density of cross-linking points and chain entanglements, that is, mesh size, which also largely determines the drug release behavior by regulating the diffusion of molecules or particulate carriers. 35 Thus, we next investigated the tunability of G 0 of the newly synthesized HA hydrogels. Changing the DC of HA-CBT and length of Cys-PEG n -Cys linkers are two straightforward ways to control the mesh size. HA-CBT with DC of 5.1 ± 0.5%, 10.3 ± 0.5%, and 18.5 ± 4.4% were synthesized and cross-linked by Cys-PEG 46 -Cys at 20 mg/ml in saline. As expected, the resulting hydrogels show increased G 0 with an increase of DC, and a maximum G 0 value of around 1 kPa was obtained at DC of $18.5%

| Evaluation of HA hydrogels for sustained drug delivery
To evaluate their potential in drug delivery, we first measured the mesh size of the optimized HA hydrogels based on the blob model, swelling ratio, and storage modulus (Equations S1-S3, Figure S4). 36 The estimated mesh size value was $36 nm. Such mesh size is small enough to efficiently encapsulate nanosized particles or crystals. To Similarly, 50 nm nanoparticles showed a similarly slow recovery behavior, with only 13.9% recovery after 10 min. A similar result was reported previously by Seiffert et al. which is attributed to the special translational mobilities of particles in polymer matries. 38 The ability to immobilize nanosized particles within the hydrogel network is indeed highly desirable for constructing drug nanocrystal depots.
Next, we evaluated the stability of hydrogel materials inside the synovial fluid. Rapid clearance and thus short joint residence time of the delivered substance represent the most challenging obstacle for IA therapy. 14,15 The stability of the IA-injected HA hydrogels is thus crucial for the ultimate performance of these drug crystal/HA hydrogels for IA therapy. HA molecules are fairly stable in hyaluronidasefree solutions, but are quickly catabolized in synovial fluid, which is primarily attributed to the existence of reactive oxygen species in synovial fluid. 39 In vivo degradation of HA in synovial fluid is a complex process, involving the diffusion of transition metals and their carrier proteins, and the generation of unstable reactive oxygen species, 39 which is challenging to be replicated in vitro or ex vivo. In a pilot study, after mixing HA with synovial fluid at 37 C for overnight, no detectable decrease in HA molecular weight was observed

| Design of drug nanocrystals formulations and in vitro release study
Crystalline CPT was formulated to demonstrate the sustained release behavior from such drug nanocrystal/HA hydrogel depots. The size of The evaluation of hydrogels for drug delivery. (a) The diffusion properties of the polystyrene nanoparticles inside the hydrogels were characterized using FRAP and compared with FITCdextran. (b) Schematic illustration of intra-articular injection and representative IVIS images of the injected knee joints of rats over 28 days. HA hydrogel was labeled with Alexa Fluor 647. The blue circles with a fixed diameter outlined the ROI range for the quantification of fluorescence. Color scale: 1.6 Â 10 8 -3.2 Â 10 9 Radiant efficiency unit: (p/s•cm 2 •sr)/(μW/cm 2 ). Noninjected rats were imaged at all the time points to collect the background fluorescent intensity (blue line in c). (c) The joint residence time of fluorescent labeled HA hydrogels. Data points were fit by a onephase exponential decay models. Each value represents the mean ± SEM (n = 5) and statistical analysis by the nonparametric Wilcoxon matched-pairs signed ranks test (*p < 0.05) drug crystals was controlled within the submicron range by manipulating the process parameters such as drug concentration, solvent/ antisolvent ratio, temperature, and precipitation time. The size and morphology of drug nanocrystals were characterized by dynamic light scattering (DLS) and scanning electron microscopy (SEM). Figure 5a,b and Figure S6 show the typical size distribution and morphology of the CPT drug crystals. These crystals were suspended in water, mixed with HA-CBT solution, and cross-linked by the addition of Cys-PEG 46 -Cys solutions. Their injectability after encapsulation was verified ( Figure 5c).
We firstly evaluated the dissolution rate of CPT nanocrystals by incubating the CPT nanocrystals in phosphate-buffered saline (PBS) at 37 C. A slow and sustained dissolution behavior was observed with around 80% of CPT dissolved over 28 days. A similar sustained drug release was observed with the CPT nanocrystal-encapsulated hydrogels (Figure 5d,e). We also found that the drug nanocrystals with size in the range between 160 and 560 nm ( Figure S6) have a similar dissolution rate either in naked or encapsulated forms, which gives flexibility for the nanocrystal preparation.
To validate the slow release for IA therapy, dissolution was also examined in RA synovial fluid (RASF) to simulate the inflamed joint environment. Compared with the release in PBS, CPT has a slower release rate with around 70% released within 28 days (Figure 5e). This could be attributed to the high viscosity of RASF in the release medium which hindered the free drug diffusion out of the dialysis membrane. The slow drug release rate further highlights the potential of such hydrogel/drug nanocrystals formulation for IA-localized drug delivery.

| Therapeutic efficacy in an arthritic rat model
Having shown the long-term stability of HA hydrogels and a sustained release of the encapsulated drug nanocrystals, we then validated the potential of CPT nanocrystal-loaded HA hydrogel for IA therapy. To this end, a collagen-induced arthritis (CIA) model was developed by intradermal injection of complete Freund's adjuvant/collagen II emulsion to SD rats. [40][41][42] The incidence of arthritis was confirmed by the erythema and swelling of the paw joints ( Figure S7). Three types of formulations were prepared, including CPT nanocrystal-encapsulated HA hydrogels (HA-CPT), blank HA hydrogels (HA), and free CPT nanocrystals (CPT) with saline as the placebo. On Day 28 of arthritis induction, each formulation was injected intra-articularly (Figure 6a).
Treatment with HA-CPT reduced arthritis severity compared to those treated by blank HA hydrogels or saline (Figure 6b). Free CPT nanocrystals did not show a significant reduction of joint size compared to saline-treated animals, which could be attributed to the rapid joint clearance of the sub-micron-sized agents. 15 The local therapeutic However, in the HA-CPT-treated group, the bone and cartilage F I G U R E 6 Therapeutic efficacy in collagen-induced arthritis animal models. (a) Experimental outline: arthritis was induced by intradermal injection of complete Freund's adjuvant/collagen II emulsion on day 0. On day 28, CPT nanocrystal encapsulated HA hydrogels (HA-CPT), blank HA hydrogels (HA), free CPT nanocrystals (CPT), and saline was injected and joint size was monitored for another 4 weeks. Animals were sacrificed on day 60 with joints being harvested. (b) Change in joint size was measured by clippers. All data are presented as mean ± SEM (n = 5) and statistical analysis by t test (*p < 0.05). (c) The relative amount of IL-1β in joint homogenates. All data are presented as mean ± SEM (n ≥ 4) and statistical analysis by one-way ANOVA (*p < 0.05, **p < 0.01, ***p < 0.001). Histological analysis with H&E staining (d, i, n) and microcomputed tomography evaluation with representative reconstructed three-dimensional images (e, j, o), coronal images (f, k, p), sagittal images (g, l, q), and X-ray projection images (h, m, r) of dissected joint tissues of healthy control (d-h), saline-treated group (i-m), and HA-CPTtreated group (n-r) (scale bar = 100 μm) An attempt to extract and quantify the plasma CPT concentrations at different time points from Day 1 to Day 32 after treatment was made by using an ultrahigh-performance liquid chromatography (HPLC)/mass spectrometer with 7-ethyl-10-hydroxyCPT as internal standard (LOD of 6 ng/ml and recovery of $76.87 ± 2.10%), but no free CPT was detected. This could be attributed to the slow dissolution rate of the CPT nanocrystals and the rapid plasma clearance rate, giving an extremely low plasma CPT concentration. This indeed is highly desirable for localized IA therapy as such low plasma CPT concentration minimizes the systemic toxicity. Histology of the heart, lungs, liver, kidney, and spleen are normal in all groups as shown in Figure S10. No body-weight reduction was observed for any of the treated animals ( Figure S11).

| DISCUSSION
There is an unmet medical need for developing new treatments for inflammatory arthritis (e.g. RA). Therapeutics that can inhibit the tumor-like aggressive proliferation of RA synoviocytes are potentially effective in treating RA. 9 However, the use of antiproliferative drugs, commonly used for cancer treatments, to treat RA is greatly limited by their systemic toxicity and low joint bioavailability. 14 CPT, as chosen here, is one such antiproliferative agent with antiarthritic potentials. 12,13 To overcome the formulation and delivery challenges, we formulated CPT into nanocrystals, encapsulated them into an injectable HA hydrogel, and intra-articularly injected it into the inflamed joints to achieve localized and sustained drug delivery.
We developed an in situ click-cross-linkable, injectable HA hydrogel using a biocompatible click reaction: CBT/cysteine reaction, which is the last step in the generation of D-luciferin in fireflies. 30 We optimized the gel modulus, a key parameter related to its internal structure and thus important for regulating drug crystal release, by tuning the degree of modification of CBT, cysteine linker size and polymer concentration. Remarkably, the HA hydrogel can reside inside the rat joint for more than 4 weeks with $35% of hydrogels remaining on day 28, which is much longer comparing to the uncross-linked HA polymer ($12 to 24 h 14 ).
The HA hydrogel described here can encapsulate CPT nanocrystals in a simple admix manner and is readily injectable to form a drug depot locally. In vitro release study confirmed that CPT can slowly release in either PBS or RASF for 4 weeks. The promise of this CPT nanocrystal/HA hydrogel formulation in achieving long-termlocalized IA drug delivery was further highlighted in vivo using a CIA model in rats. Animals treated with the experimental formulation showed a reduction in arthritis severity.
It is important to note the limitations of this study. In this work, we did not detect any circulating CPT molecules in the plasma at different time points, thus future animal studies should incorporate experiments to evaluate the plasma and joint pharmacokinetic profiles using radiolabeled drugs. Systemic biodistribution of these radiolabeled drugs could also be studied. Also, rats were sacrificed for efficacy and toxicity evaluation at Day 60 and found no signs of organ toxicity. It would be necessary to evaluate the histotoxicity and cartilage destruction at different time points.
In summary, we have developed an effective drug nanocrystalencapsulated HA hydrogel system for IA therapy. This drug nanocrystal/

| CPT nanocrystal preparation
CPT nanocrystals were prepared using an antisolvent precipitation method under sonication, as reported previously. 43

| Hydrogel fabrication with or without drug nanocrystals
All hydrogels were prepared in a similar manner. For a typical procedure of blank hydrogels without drug nanocrystals, HA-CBT with DC of 10% was dissolved in saline at a concentration of 20 mg/ml and mixed thoroughly with freshly prepared Cys-PEG 46 (1).
Similarly, Cys-PEG-Cys linkers and their protected forms were dissolved in D 2 O and acetone-d 6 , respectively, and their chemical shifts were reported with reference to the solvent peak.

| Gel permeation chromatography
HA-CBT conjugates were analyzed by a Viscotek 270max GPC system

| Rheological characterization
The rheological properties of hydrogels were measured by an AR-G2

| IVIS imaging
Alexa Fluor 647-labeled HA was used for the fabrication of HA hydrogels. After IA injection, an IVIS (Spectrum, PerkinElmer, USA) with an anesthesia system and Living Image software (3.1, Caliper, USA) was used to image and quantify the fluorescence of joint over 28 days.

| Joint homogenization and protein retrieval
After euthanizing, treated joints with surrounding synovial tissues were dissected and frozen in liquid nitrogen, pulverized by a mortar and pestle, and homogenized in a lysis buffer (89901, Thermo Scientific, USA) containing protease and phosphatase inhibitors (78442, Thermo Scientific, USA) and centrifuged for 10 min at 400 g (4 C) to collect the supernatants. IL-1β concentration was quantified by ELISA (BMS630, Thermo Scientific, USA).

| Microcomputed tomography
Animal knee joints were analyzed by micro-CT using the Nikon

| Histological analysis
Retrieved samples were fixed in 10% neutral buffered formalin, embedded in paraffin, and sectioned according to standard histological process. Slides were then stained with hematoxylin and eosin (H&E) and safranin-O.

| Statistics analysis
All data were reported as means ±SEM from separate experiments. Statistical analyses were performed by t-test or one-way ANOVA methods using GraphPad Prism (GraphPad Software, USA). Differences were considered statistically significant when p < 0.05.