Abstract
Biologic drug therapies are effective treatments for autoimmune diseases such as rheumatoid arthritis (RA) but may cause significant unwanted adverse effects, as they are administered continuously at high doses that can suppress the immune system. As the severity of RA fluctuates over time, targeted strategies that can dynamically sense and respond to changing levels of endogenous inflammatory mediators may achieve similar therapeutic efficacy while reducing risks of adverse effects. Using CRISPR-Cas9 genome editing, we engineered stem cells that harbor a synthetic gene circuit expressing biologic drugs to antagonize interleukin-1 (IL-1) or tumor necrosis factor α (TNF-α) in an autoregulated, feedback-controlled manner in response to activation of the endogenous chemokine (C-C) motif ligand 2 (Ccl2) promoter. To examine the in vivo therapeutic potential of this approach, cells were tissue-engineered to form a stable cartilaginous scaffold, which was implanted subcutaneously in mice with inflammatory arthritis. These bioengineered anti-cytokine implants mitigated arthritis severity as measured by joint pain, structural damage, and systemic and local inflammation. The coupling of synthetic biology with tissue engineering promises a wide range of potential applications for treating chronic diseases by generating custom-designed cells that regulate the expression of therapeutic transgenes in direct response to dynamically changing pathologic signals in the body.
Despite advances in the development of disease-modifying anti-rheumatic biologic drugs, approximately 40% of patients with rheumatoid arthritis (RA) fail to respond to treatment1. And while the severity of RA fluctuates, biologic drugs are administered continuously at high concentrations, predisposing patients to significant adverse effects, such as increased risk of infection2. The development of therapeutics that can sense and respond to dynamically changing levels of endogenous inflammatory mediators may improve efficacy while mitigating the side effects of continuous biologic delivery3–5. Here, we used CRISPR-Cas9 genome engineering6,7 to create a self-regulating gene circuit in induced pluripotent stem cells (iPSCs). These cells were designed to produce anti-cytokine biologic drugs in response to inflammatory signals such as interleukin-1 (IL-1) or tumor necrosis factor a (TNF-α) by transcribing their biologic inhibitors in a feedback-controlled manner3, driven by the promoter of the chemokine (C-C) motif ligand 2 (Ccl2) (Fig. 1)8. For delivery in vivo, iPSCs were differentiated into a cartilaginous implant that maintains the cells in a stable subcutaneous depot, allowing for sensing of systemic inflammation as well as free diffusion of the biologic drugs into the circulation. We demonstrated that these bioengineered implants provide dynamic, autoregulated delivery of anti-cytokine biologic drugs that mitigate structural damage, pain, and inflammation in a murine model of arthritis.
Using CRISPR-Cas9 genome engineering, iPSCs were edited to insert either the gene for IL-1 receptor antagonist (Il1rn) or luciferase (Luc) at the Ccl2 locus, creating a self-regulating gene circuit that transcribes Il1rn or Luc in response to inflammatory activation of Ccl2 (referred as Ccl2-IL1Ra or Ccl2-Luc cells)3. We initially examined the therapeutic potential of Ccl2-Luc cells9 by intraperitoneal (IP) injection into a murine model of K/BxN serum transfer arthritis (STA)10. When triggered exogenously by a single IP injection of TNF-α to simulate an RA “flare”, we observed robust luciferase expression (Suppl. Fig. 1a) that wanes after the first 24 h (Suppl. Fig. 1b). In the context of K/BxN STA, Ccl2-IL1Ra cells delivered by IP injection did not mitigate clinical scores (p=0.10, Suppl. Fig. 1c), but significantly suppressed ankle swelling (13.8% reduction on day 6) and mitigated pain sensitivity (25% reduction in untreated animals, no change in IL-1Ra treated animals; Suppl. Fig. 1d-f).
Due to the lack of long-term cell engraftment by IP injection (as measured by luciferase expression after 24h), as well as the modest mitigation of disease severity, we used an alternative tissue-engineering approach to create a cartilaginous construct to provide stable engraftment. Cells were seeded onto 3D woven poly(ε-caprolactone) (PCL) scaffolds (Fig. 2a)11 and chondrogenically differentiated12 over 21 days into chondrocyte-like cells that produced a proteoglycan-rich matrix (Fig. 2b-d). In culture, the Ccl2-IL1Ra bioartificial implants sensed and responded to stimulation with IL-1α by producing IL-1Ra (Fig. 2e) in a feedback-controlled manner (Fig. 2f,g). A second set of cells were similarly engineered to produce soluble TNF receptor 1 (sTNFR1), and implants created from Ccl2-sTNFR1 cells behaved similarly in vitro (Suppl. Fig. 2). The cell-secreted anti-cytokine biologics suppressed the response to IL-1α or TNF-α, reducing mRNA levels for proinflammatory mediators Il6 and Ccl2, and for matrix metalloproteinases, Mmp9 and Mmp13 (Fig. 2h-k, Suppl. Fig. 2), as compared to the control Ccl2-Luc construct.
These bioartificial constructs were implanted subcutaneously onto the dorsum of K/BxN transgenic F1 mice, which develop spontaneous arthritis, as well as mice with K/BxN STA10. Ccl2-Luc constructs exhibited high luminescence relative to background levels, indicating responsiveness of the constructs in vivo, as well as cell survival in a chronic inflammatory environment up to 5 weeks after implantation in F1 K/BxN mice (Fig. 2l,m). In the K/BxN STA model Ccl2-Luc constructs exhibited similar luminescence signals for 7 days when they were implanted subcutaneously near the knee or on the dorsum of mice (Fig. 2n). Explanted Ccl2-Luc constructs from F1 K/BxN mice at 5 weeks also confirmed high cell viability via live/dead staining by confocal microscopy (Fig. 2o).
Chondrocyte-like cells in a cartilage matrix exhibit minimal migration and require no vasculature for long-term survival, relying on diffusion for nutrient transport13–16. To determine the ability of this selfregulating implant to mitigate RA disease severity, constructs were implanted on the dorsal aspect of C57BL/6 mice and allowed to recover for 7 days followed by intravenous K/BxN serum administration on day 0 (Fig. 3a). Animals receiving Ccl2-IL1Ra implants demonstrated significant improvements in RA severity, including 40% reductions in both clinical scores and ankle thickness, compared to control animals that received Ccl2-Luc constructs or no implant (Fig. 3b,c). Histologic analysis showed that animals with Ccl2-IL1Ra implants demonstrated significantly lower inflammation scores and had reduced cartilage degradation and proteoglycan loss (Fig. 3 d-f) when compared to animals with Ccl2-Luc implants. Mice receiving Ccl2-IL1Ra constructs also exhibited significantly higher serum levels of IL-1Ra when compared to controls (Fig. 3g,h). Serum IL-1Ra concentrations correlated with clinical scores and ankle thickness (Fig. 3i,j). The reduction in disease severity in mice with Ccl2-IL1Ra constructs was accompanied by a significant decrease in mechanical pain sensitivity as measured by algometry and Electronic Von Frey pain tests, whereas the control groups exhibited increased pain sensitivity (reduced pain thresholds) (Fig. 3k,l). The pain threshold of each animal also significantly correlated with serum levels of IL-1Ra (Fig. 3m,n). Neither Ccl2-sTNFR1 (Suppl. Fig. 3) nor reducing the dose of Ccl2-IL1Ra constructs by half (Suppl. Fig. 4) significantly mitigated structural damage from K/BxN STA, although both approaches significantly mitigated an increase in pain sensitivity (Suppl. Fig. 3, 4). These data corroborate previous reports that TNF-α and IL-1α signaling are both directly and indirectly associated with the onset of pain in K/BxN STA17.
Microcomputed tomography (microCT) of bone structure showed that mice receiving Ccl2-IL1Ra implants had minimal or no bone erosions on day 7 following K/BxN serum transfer, whereas mice with Ccl2-Luc implants or no implant exhibited highly erosive disease as demonstrated by a higher ratio of bone surface to bone volume (BS/BV) and a lower ratio of bone volume per total volume (BV/TV) (Fig. 4a-c). Histologic analysis was also concordant with the microCT observations for bone erosions (Fig. 3d). Ccl2-IL1Ra implants also showed protective effects on bone mineral density (BMD) (Fig. 4d). Quantitative measures of bony erosions (BS/BV) negatively correlated with serum levels of IL-1Ra (Fig. 4e). Bony erosions are frequently observed in RA18 and are associated with disease severity and poor functional outcome. However, most biologics currently available in the clinic rarely inhibit both bone erosions and inflammation18. The ability to mitigate both inflammation and structural damage using this approach suggests that inhibition of IL-1 signaling in an autoregulated manner may provide added therapeutic advantage in the prevention of bone loss during inflammatory arthritis. Finally, we compared the clinical response of animals with K/BxN STA receiving Ccl2-IL1Ra implants to those treated with methotrexate, a conventional disease-modifying drug, or tofacitinib, a JAK-STAT biologic inhibitor 19, and saw no significant mitigation of arthritis, structural changes, pain outcomes, or bony erosions with these treatments (Suppl. Fig 5).
Mice receiving Ccl2-IL1Ra implants showed significant reductions in their inflammatory cytokine profiles as compared to control animals. Using a multiplexed cytokine assay, we observed significant reductions of serum levels for IFN-γ and IL-6, and an increase in serum levels for IL-10 in treated animals (Fig. 4f-h). In the paw lysate of treated animals, we observed significant reductions in levels for IL-1α, IL-6, MCP-1, KC, and TNF-α in comparison to control animals (Fig. 4i-m). There was a trend toward significant reduction in total level of IL-1 (IL-1α + IL-1β) in animals that received Ccl2-IL1Ra implants compared to controls (p=0.07, Fig. 4n).
In summary, we developed a genome-engineered implantable drug delivery system that can automatically sense and respond to inflammatory cytokines to produce therapeutic levels of anti-cytokine biologics in an auto-regulated manner. By engineering iPSCs to form a cartilaginous matrix, the implants exhibited long-term viability and function in vivo, with the ability to serve as a continuous source for biologic drugs. This platform represents a clinically-relevant and translational delivery strategy that may be applicable to a wide variety of diseases for long-term therapeutic control. While protein-based drug delivery, including enzyme-activated systems, have been successful in a number of applications20–23, such approaches are generally based on biodegradable materials and cannot easily supply continuously biologic molecules for long-term delivery, unlike the rapid dynamic stimulus-response functions that living cells can provide. In previous studies, “designer” or “smart” cell approaches have been used to develop cell therapies for metabolic diseases such as diabetes, but have been based on transient transfection or viral gene delivery, resulting in unpredictable gene insertion sites or copy numbers4. This study is among the first to demonstrate use of CRISPR-Cas9 gene editing for precise control of the stimulus-response locus in vivo3, as well as the potential for tunability and genome editing of cytokine receptors for treating model systems of an inflammatory condition24,25. Applying tissue engineering in combination with synthetic biology26 to develop artificial gene circuits that can sense and respond dynamically to disease markers may provide new opportunities for developing safe and effective therapies for chronic autoimmune diseases such as RA27.
Methods
Cell Culture and In Vitro Assays
Murine induced pluripotent stem cells (iPSCs) were engineered to incorporate a synthetic gene circuit to drive the expression of either Il1rn or firefly luciferase (Luc) via the Ccl2 locus3. These iPSCs were predifferentiated in micromass with BMP-412, seeded onto 3D woven poly(ε-caprolactone) (PCL) scaffolds (1e6 cells/scaffold), and cultured for 21 days in serum-free chondrogenic medium containing Dulbecco’s Modified Eagle Medium - High Glucose (DMEM-HG), non-essential amino acids, 2-mercaptoethanol, ITS+ premix (BD), penicillin-streptomycin (Gibco), 50 μg/mL L-ascorbic acid 2-phosphate, 40 μg/mL L-proline, 10 ng/mL transforming growth factor beta 3 (TGF-β3, R&D Systems) and 100nM dexamethasone12,28 to make implantable cartilage constructs11. Gene expression changes in response to inflammatory cytokines (1 ng/mL IL-1α or 20 ng/ml TNF-α) were measured via qPCR (primers available in Supplementary Methods), and protein levels of secreted murine IL-1Ra were measured via ELISA (Duoset, R&D Systems).
nanoCT Imaging of Bioartificial Implants
Implants were fixed overnight in 2.5% glutaraldehyde and 2% paraformaldehyde in 0.1 M cacodylate buffer with 2 mM CaCl2. Secondary fixation was performed using 1% OsO4 in 0.1 cacodylate buffer for 1 hour at room temperature. Excess fixative was washed using deionized water, and the samples were incubated in Lugol’s iodine for 72 hours at room temperature to further enhance contrast. Samples were embedded in 2% agarose and imaged on a Zeiss Xradia Versa 540 X-ray microscope using a 0.4x flat panel detector. Images were reconstructed and presented as shaded volume renderings.
In Vivo Luciferase Assays
Ccl2-Luc constructs were implanted on the dorsal aspect of F1 K/BxN mice, which spontaneously develop chronic arthritis, to allow for longitudinal bioluminescence imaging of luciferase activity of the scaffolds in vivo daily for one week, and weekly thereafter over a 5-week period. C57BL/6 mice implanted with Ccl2-Luc scaffold constructs were challenged with K/BxN serum and imaged daily for 7 days (Supplementary Methods).
Cell Viability Assays
Cell survival and apoptosis were assessed using the Live/Dead® Cell Viability/ Cytotoxicity Kit for mammalian cells (Invitrogen/Molecular Probes, Carlsbad, California, USA). Live cells were labeled with calcein AM and dead cells were labeled with ethidium homodimer-1 bound to DNA. Labeled constructs were imaged using confocal microscopy (LSM 880, Zeiss, Thornwood, NY, USA)11.
K/BxN Model of Inflammatory Arthritis
Male K/BxN transgenic mice and female B6.I-Ag7 mice were intercrossed to generate F1 that spontaneously developed arthritis beginning at about 4 weeks of age and lasting for > 20 weeks10. STA was induced in C57BL/6 male and female mice with 150-175μL of K/BxN serum delivered by retroorbital injection10,29,30. One week prior to STA, mice received either subcutaneous Ccl2-IL1Ra scaffold implants on the dorsum (n=8 scaffolds, 8 animals, Experimental group), Ccl2-Luc scaffolds or no treatment (Control groups, n=8-13). In separate groups of C57BL/6 mice, animals were induced with STA and received methotrexate IP, tofacitinib by oral gavage, or appropriate controls as detailed in Supplementary Methods. Disease activity (clinical score and ankle thickness) and pain sensitivity (algometry and Electronic Von Frey) were assessed daily for 1 week. Mice were then sacrificed, and serum and hindpaws were collected for analysis. All procedures were approved by IACUC at Washington University in St. Louis.
Histological Analysis
Paws were harvested on day 7 after serum transfer, fixed in 10% neutral buffered formalin for 48 h and stored in 70% ethanol, before decalcification in EDTA solution and processing for paraffin embedding. Sections (5 μm) were stained with Hematoxylin and Eosin (H&E) or toluidine blue. Inflammatory cells infiltrating the synovial lining and the joint cavity were enumerated in 8-10 random fields per section using H&E images acquired at 400x magnification. Proteoglycan loss in the cartilage was scored on toluidine blue-stained sections on a scale from 0–3, ranging from fully stained cartilage (score = 0) to fully destained cartilage (score = 3) as previously described31. Scoring was performed by an observer blinded to the treatment.
Quantitation of IL-1Ra and Proinflammatory Mediators
Levels of IL-1Ra in serum and paws were assessed by ELISA (Quantikine-IL1Ra, R&D Systems). Levels of paw and serum inflammatory mediators were evaluated using Luminex® (Mouse High Sensitivity T-cell discovery array 18-Plex, Eve Technologies, Calgary, AB, Canada).
MicroCT Analysis of Bone Erosion
To measure bone morphological changes, intact hind paws were scanned by microcomputed tomography (microCT, SkyScan 1176, Bruker) with a 9 μm isotropic voxel resolution at 50 kV, 500 μA, 980 ms integration time, 3 frame averaging, and 0.5 mm aluminum filter to reduce the effects of beam hardening. Images were reconstructed using NRecon software (with 20% beam hardening correction and 15 ring artifact correction). Hydroxyapatite calibration phantoms were used to calibrate bone density values (g/cm3). Parameters reported are: bone surface to bone volume (BS/BV), bone fraction (bone volume/total volume; BV/TV), and bone mineral density (BMD; g/cm3).
Statistical Analysis
Sample size was determined based on a mean clinical score of 10 ± 2 for KRN control animals and 6 for IL-1Ra treated animals. Based on an alpha of 0.05 and 80% statistical power (1-β) a priori a sample size of 4 animals/treatment group was needed to observe this effect. Outcomes were evaluated by two-way Student’s t-test, one-way Mann-Whitney U test or one-, two-, three-way repeated measures, or three-way ANOVA with Tukey’s post hoc test or Sidak Correction to assess differences between groups, treatments, scaffold types, time, or a combination of those factors. Pearson or Spearman correlations were calculated between serum levels of IL-1Ra and outcomes (p ≤ 0.05). Investigators were not blinded to the clinical score or ankle thickness measurements. All other assessments and analyses were performed blinded.
Author Contributions
Y.R.C., K.H.C., C.T.N.P., and F.G. conceived the project, Y.R.C., K.H.C., C.L.W., J.M.B., C.T.N.P. and F.G. designed the experiments. Y.R.C., L.P., A.K.R., F.T.M., and C.L.W. conducted in vitro studies. In vivo experiments and data analyses were performed by Y.R.C, K.H.C., L.E.S., L.P., A.K.R., N.S.H., and C.L.W. Y.R.C., K.H.C., and F.G. wrote the manuscript. All the authors read, edited, and approved the final manuscript.
Data Sharing
The data that support the findings of this study are available from the corresponding author upon reasonable request.
Acknowledgments
We thank Dr. Charles Gersbach for important discussions in the early stages of this work and Sara Oswald for providing technical writing support for the manuscript. This study was supported in part by grants NIH grants AR50245, AR48852, AG15768, AR48182, AG46927, OD10707, DK108742, AR073752, AR057235, AR067491, the Arthritis Foundation, and the Nancy Taylor Foundation for Chronic Diseases, NIH P50 CA094056 (Molecular Imaging Center) and NCI P30 CA091842 (Siteman Cancer Center Small Animal Cancer Imaging shared resource).