Research ArticleNanomedicine

Cartilage-penetrating nanocarriers improve delivery and efficacy of growth factor treatment of osteoarthritis

See allHide authors and affiliations

Science Translational Medicine  28 Nov 2018:
Vol. 10, Issue 469, eaat8800
DOI: 10.1126/scitranslmed.aat8800

Dendrimer deliverance for cartilage degeneration

There are currently no disease-modifying drugs to treat osteoarthritis (OA), a debilitating disease affecting the joints. Here, Geiger et al. developed dendrimer-based nanocarriers to deliver insulin-like growth factor 1 (IGF-1) to chondrocytes within joint cartilage. By tuning the surface charge of the nanocarriers, dendrimer–IGF-1 could penetrate full-thickness bovine cartilage ex vivo. Nanocarriers injected into joints were retained in rat knees in models of OA and reduced cartilage degeneration. This study suggests that targeting delivery of disease-modifying agents to chondrocytes using dendrimer nanocarriers could be therapeutic for OA.


Osteoarthritis is a debilitating joint disease affecting nearly 30 million people for which there are no disease-modifying therapies. Several drugs that have failed clinical trials have shown inefficient and inadequate delivery to target cells. Anabolic growth factors are one class of such drugs that could be disease-modifying if delivered directly to chondrocytes, which reside deep within dense, anionic cartilage tissue. To overcome this biological barrier, we conjugated a growth factor to a cationic nanocarrier for targeted delivery to chondrocytes and retention within joint cartilage after direct intra-articular injection. The nanocarrier uses reversible electrostatic interactions with anionic cartilage tissue to improve tissue binding, penetration, and residence time. Amine terminal polyamidoamine (PAMAM) dendrimers were end functionalized with variable molar ratios of poly(ethylene glycol) (PEG) to control surface charge. From this small family of variably PEGylated dendrimers, an optimal formulation showing 70% uptake into cartilage tissue and 100% cell viability was selected. When conjugated to insulin-like growth factor 1 (IGF-1), the dendrimer penetrated bovine cartilage of human thickness within 2 days and enhanced therapeutic IGF-1 joint residence time in rat knees by 10-fold for up to 30 days. In a surgical model of rat osteoarthritis, a single injection of dendrimer–IGF-1 rescued cartilage and bone more effectively than free IGF-1. Dendrimer–IGF-1 reduced width of cartilage degeneration by 60% and volumetric osteophyte burden by 80% relative to untreated rats at 4 weeks after surgery. These results suggest that PEGylated PAMAM dendrimer nanocarriers could improve pharmacokinetics and efficacy of disease-modifying osteoarthritis drugs in the clinic.


Osteoarthritis (OA) is a debilitating disease of individual joints that manifests as degeneration of articular cartilage, causing pain and impeding mobility. It affects 20 to 30 million people in the United States alone with an estimated economic impact of $60 billion/year (1, 2). Despite this enormous unmet medical need, no disease-modifying drug exists, and the current standard of care is limited to palliative treatment to manage symptoms as the disease progresses. Although several classes of drugs have shown promise in preclinical studies—including anti-inflammatory small molecules, cytokine receptor antagonists, anabolic growth factors, and targeted inhibitors of catabolic enzymes—all have failed in clinical development (3). Even the current pharmacological standard of care for pain (specifically intra-articular corticosteroids and hyaluronic acid) is subject to intense debate regarding its safety and efficacy (4, 5). Many of these shortcomings are rooted in inadequate drug delivery (6, 7).

The rapid clearance rate of molecules from the joint space after direct injection into the affected joint and the dense, avascular nature of cartilage tissue constitute two considerable biological barriers to drug delivery to chondrocytes. Many compounds of interest have intra-articular half-lives as short as 2 to 4 hours (7, 8). Combined with a need for clinicians to minimize repeat intra-articular injections, the short half-life of these drugs limits their duration of therapy and thus overall efficacy. In a retrospective clinical analysis (NCT00110916), Chevalier et al. (9) identified insufficient drug delivery as a key factor in the failure of an interleukin 1 receptor antagonist (IL1-RA; anakinra), an approved disease-modifying rheumatoid arthritis drug, in a phase 2 trial for OA. Patients received a single injection of IL1-RA over a 12-week study, yet the drug was below its limit of quantification in serum after 24 hours (9).

Previous studies have shown that small (<15 nm) cationic nanocarriers can overcome the biological barriers of the joint by binding and penetrating anionic cartilage tissue faster than the carriers can be cleared from the joint space (1016). It is vital for clinical translation of these cartilage-penetrating nanocarriers to identify an optimal nanocarrier surface charge that is both safe and efficacious. Beyond small size and tunable surface charge, other important considerations for a translatable, cartilage-penetrating nanocarrier include scalable synthesis, robust characterization, and flexibility in accommodating different classes of therapeutic cargo. We found that polyamidoamine (PAMAM) dendrimers, generation (Gen) 4 to Gen 6, fit all design criteria well. These hierarchically branched macromolecules are <10 nm in diameter with a dense surface functionality of 64 to 256 cationic primary amines. These amines can be modified with poly(ethylene glycol) (PEG) oligomers at nearly stoichiometric conversions for tight control of charge and improved biocompatibility (1720). The surface amines also provide a versatile chemical handle for conjugation to a variety of drugs (19, 21). PAMAM dendrimers are produced commercially at large scale and can be thoroughly characterized by well-established methods (18, 22, 23).

We chose to use insulin-like growth factor 1 (IGF-1) as our therapeutic cargo. IGF-1 is an anabolic growth factor that promotes chondrocyte survival, proliferation, and biosynthesis of cartilage matrix macromolecules (2426). It also shows anti-inflammatory effects in cytokine-challenged cartilage tissue (24, 27). Because of these properties, IGF-1 has garnered considerable interest as a potential disease-modifying OA drug.

Our goal was to engineer a family of cartilage-penetrating nanocarriers with variable surface charge, select an optimally charged system, and rigorously test the selected system for improved drug delivery and efficacy in vitro and in vivo. We identified an optimally charged, PEGylated dendrimer capable of high uptake into cartilage tissue without toxicity and conjugated it to IGF-1 without loss in bioactivity. The conjugate increased residence time in the joint compared to free IGF-1. In a rat surgical model of OA (2831), a single intra-articular injection of dendrimer–IGF-1 reduced cell and aggrecan loss relative to untreated rats at 4 weeks after surgery, whereas free IGF-1 did not. This study provides evidence that conjugation to a cartilage-penetrating PEGylated dendrimer can improve both delivery and efficacy of disease-modifying biologic drugs in OA.


Dendrimer–IGF-1 conjugation retains IGF-1 bioactivity

PAMAM dendrimers are densely cationic macromolecules able to bind to anionic cartilage tissue. We hypothesized that cartilage-bound dendrimers would exhibit reduced clearance from the synovial fluid through joint microvasculature (Fig. 1A). The small size of dendrimers could enable penetration into dense cartilage to access IGF-1 receptors on chondrocytes throughout the entirety of the tissue and the presence of partial PEG shielding of surface charge could enable dynamic binding-unbinding interactions favoring deep penetration into cartilage over surface binding. Gen 4 (14 kDa, 64 NH2 end groups) and Gen 6 (58 kDa, 256 NH2 end groups) dendrimers were used (Fig. 1B, Gen 4 shown), and 0 to 60% of end groups were PEGylated to covalently and sterically shield surface charge. This method created a small panel of variably charged formulations for testing in model systems. Remaining unreacted primary amine end groups contributed to positive surface charge.

Fig. 1 Design and synthesis of bioactive PEGylated dendrimer–IGF-1 conjugates as cartilage penetrating nanocarriers.

(A) Schematic of drug and carrier fates within joints after intra-articular injection based on size and charge. Material within the synovial fluid is rapidly cleared from the joint. Cationic material binds to anionic aggrecan within cartilage to avoid clearance. Small cationic materials can penetrate the aggrecan matrix to interact with chondrocytes deep in the tissue. ECM, extracellular matrix. (B) Chemical structure and key design characteristics of PAMAM dendrimers (Gen 4 shown) as cartilage-penetrating nanocarriers. (C) Synthetic scheme of a partially PEGylated dendrimer–IGF-1 formulation. For the Gen 4 35% PEG formulation, n = 64, x ≅ 18, y ≅ 4.4, and z ≅ 0.6. For the Gen 6 45% PEG formulation, n ≅ 224, x ≅ 90, y ≅ 9.4, and z ≅ 0.6. IGF-1 color code: orange, IGF-1 receptor binding site (Phe-Tyr-Phe); blue, lysine (alternative reaction site); cyan, matrix IGF-1 binding protein site (hinders IGF-1 transport); magenta, fluorescent tracer. (D) Rate of biosynthetic 35S sulfate incorporation from media into ex vivo bovine cartilage after treatment with no IGF-1, dendrimer–IGF-1, or free IGF-1. Data are means + 95% confidence interval (CI), n = 10 explants, and statistics by one-way analysis of variance (ANOVA) with Tukey post hoc tests. sGAG, sulfated glycosaminoglycan; ww, wet weight. (E) Percentage of NIH/3T3 cells in S phase after 24 hours of treatment with no IGF-1, dendrimer–IGF-1, or free IGF-1, as determined by cell cycle flow cytometry with 4′,6-diamidino-2-phenylindole and 5-ethynyl-2′-deoxyuridine staining. Data are means + 95% CI, n = 6 biological replicates, and statistics by one-way ANOVA with Tukey post hoc test. (F) Confocal image of interaction of dendrimer–IGF-1 conjugates with live bovine cartilage tissue. Binding to extracellular matrix and dense binding to cellular membrane (white arrows) are visible.

The dendrimer–IGF-1 bioconjugation scheme (Fig. 1C) resulted in a nondegradable thiol-maleimide linkage. The dendrimer–IGF-1 conjugate retained equivalent bioactivity to free IGF-1, both inducing sulfated proteoglycan synthesis in bovine cartilage explants (Fig. 1D) and proliferation in NIH/3T3 fibroblasts (Fig. 1E and fig. S1). Confocal microscopy of sections of bovine cartilage tissue explants incubated with fluorescent dendrimer–IGF-1 revealed colocalization of dendrimer–IGF-1 with cartilage tissue and cellular membranes (Fig. 1F). On the basis of the combined evidence of Fig. 1 (D to F), the site of membrane localization is likely the extracellular IGF-1 receptor, although some nonspecific electrostatic interaction with the membrane is also possible.

Screening identifies optimally PEGylated dendrimers

PEGylation of the small panel of Gen 4 and Gen 6 dendrimers was quantified by 1H nuclear magnetic resonance (NMR) spectroscopy and matrix-assisted laser desorption/ionization–time of flight (MALDI-TOF) mass spectrometry as a percentage of total end groups (Fig. 2A, figs. S2 and S3, and tables S1 and S2). Dendrimer formulations are hereafter referred to as Gen AB%, where A is dendrimer generation and B is the percentage of terminal amines conjugated with PEG [number-average molecular weight (Mn) = 436 Da; xavg = 8]. The small panel of partially PEGylated dendrimers was then incubated with bovine cartilage tissue disks for 24 hours (Fig. 2, B and C). Cartilage uptake was measured as the percentage of dendrimer from a 300-μl bath within a ~7-μl explant (Fig. 2D, colored, and fig. S4). The PEGylated dendrimers were counterscreened for cell viability of human CHON-001 chondrocytes after treatment for 48 hours (Fig. 2D, black, and fig. S5). Dendrimer formulations with higher percentage of PEGylation (less charge) showed not only reduced uptake into anionic cartilage tissue but also greater cell viability. All formulations showed some preferential partitioning into cartilage greater than equilibrium diffusion into the cartilage as indicated by uptake of >2.3%. When cartilage with bound dendrimers was transferred to 10× phosphate-buffered saline (PBS), nearly 100% of dendrimers desorbed from the tissue, further suggesting an electrostatic dendrimer-cartilage binding mechanism that can be shielded in the presence of free ions in the hyperosmotic PBS (fig. S6). These data were used to select one formulation from each generation for further testing; the selected systems show the highest percentage uptake possible while maintaining 100% cell viability at 10 μM dendrimer (Fig. 2D, shaded region). Gen 4 35% PEG (42 free surface amines) and Gen 6 45% PEG (123 free surface amines; see table S1) were selected for further study, with mean uptakes of 51 and 71%, respectively, and ~100% viability.

Fig. 2 Screen identifies optimal degree of dendrimer PEGylation for enhanced electrostatic cartilage binding without toxicity.

(A) Synthesis (left) and characterization pipeline (right) of dendrimers with varying lengths and molar ratios of PEG grafted to end groups. (B) Schematic of fluorescent dendrimer uptake and viability experiments in bovine cartilage tissue explants (disks). (C) Bright-field image of fluorescent dendrimer uptake in bovine cartilage 24 hours after incubation. (D) Fluorescent dendrimer uptake in bovine tissue over 24 hours and human CHON-001 cell viability 48 hours after treatment with 10 μM partially PEGylated dendrimers (PEG Mn = 436 Da; xavg = 8). Theoretical surface charge is shown. Selected (shaded region) refers to the dendrimer formulation used for further study. Data are means + 95% CI, n = 15 explants for uptake, and n = 4 technical replicates for viability. ***P < 0.001 versus all others by one-way ANOVA with Tukey honestly significant difference (HSD) posttests. (E) Cell viability staining of sectioned bovine cartilage explants after 48 hours of incubation with PEGylated dendrimers. Cell death at edges is due to artifact from tissue harvest. Scale bars, 200 μm. (F) Histology [hematoxylin and eosin (H&E)] of rat knee synovium and cartilage 2 months after intra-articular injection of nanocarrier. Scale bars, 200 μm. B, bone; C, cartilage.

Optimally PEGylated dendrimers do not exhibit acute or chronic histotoxicity

The two formulations selected from screening were further tested for toxicity ex vivo in bovine cartilage tissue disks and in vivo in rats. PEGylated dendrimers (1 and 10 μM) were incubated with cartilage disks for 48 hours, and then, tissue sections were cut, stained for live/dead cells, and imaged (Fig. 2E and fig. S7). Dendrimer-treated tissues exhibited cell viability equivalent to or greater than vehicle control (cell death at upper and lower edges is artifact from tissue harvesting). Toxicology markers in serum samples taken 2 and 7 days after intra-articular injection of dendrimer–IGF-1 (approximate knee concentration, 10 μM dendrimer) into healthy rats were either within reported normal ranges or statistically equivalent to uninjected animals (table S3). A separate group of rats euthanized 2 months after injection showed normal histopathology of joints (Fig. 2F), liver, kidney, and lungs (fig. S8). We also tested an under-PEGylated dendrimer, Gen 4 20% PEG, that had shown mild cytotoxicity in a human chondrocyte cell line (Fig. 2D). Gen 4 20% PEG induced considerable inflammation in the synovium of the rat knee, indicating the importance of optimized surface charge (fig. S9).

Dendrimer–IGF-1 considerably extends joint residence time of drug

Fluorescently labeled Gen 4 35% IGF-1, Gen 6 45% IGF-1, and free IGF-1 were injected intra-articularly into rat knees (Fig. 3A). Concentrations were tuned to ensure equal quantity of fluorophore in each formulation, and all formulations were injected to about 6 μM IGF-1 (~10 μM dendrimer) in the synovial fluid. IGF-1 fluorescence within joints was serially measured over a period of 1 month using an in vivo imaging system (IVIS; Fig. 3B). Total radiant efficiency data within the anatomical region of interest (ROI) were quantified and plotted over time. Fitting the data to a single-phase exponential decay function, free IGF-1 had a joint half-life of only 0.41 days (95% CI, 0.31 to 0.57 days), whereas Gen 4 35% extended that half-life to 1.08 days (95% CI, 0.84 to 1.5 days), and Gen 6 45% extended joint half-life even further to 4.21 days (95% CI, 2.8 to 8.5 days; Fig. 3C). These half-lives and 95% CIs were used to calculate time at therapeutically relevant dose in the joint based on the initial injected joint concentration (6 μM IGF-1) and aggrecan biosynthesis-saturating IGF-1 concentration (0.04 μM; Fig. 3D) (24, 32). A single injection of Gen 6 45% PEG–IGF-1 provided 30.4 days at IGF-1 saturating conditions within the joint, relative to 7.8 days for Gen 4 35% PEG–IGF-1 and 2.9 days for free IGF-1.

Fig. 3 PEGylated dendrimer nanocarriers extend joint residence, therapeutic time, and cartilage penetration of IGF-1 in rats.

(A) Schematic of rat intra-articular injection and IVIS image of uninjected rat to establish background signal. The field of view for (B) is shown in the red box. (B) Representative IVIS images of rat knee joints over 28 days after injection of fluorescent IGF-1 formulations. Black circles represent the anatomical joint ROI used for quantification. Fluorescence scale: max = 9.0 × 107, min = 2.0 × 107 (units in Materials and Methods). (C) Time course of fluorescent radiant efficiency within joints. Data are fit to a one-phase exponential decay with a common plateau based on background signal. Data are means + 95% CIs of nonlinear fit, n = 8 joints per formulation. Half-lives were statistically different for each dataset (P < 0.0001) by extra sum of squares F test. (D) Estimated time at IGF-1 concentrations that saturate aggrecan biosynthesis activity for each delivery method, based on initial concentration after injection of ~6 μM IGF-1 (saturating concentration of 0.04 μM) and mean + 95% CI of carrier half-life. (E) 3D reconstruction of multiphoton microscopy images of full-thickness cartilage from intact rat femurs harvested 6 days after injection. Color code: gray, collagen II second-harmonic generation signal; red, aggrecan antibody; blue, IGF-1.

Dendrimer–IGF-1 penetrates the full thickness of cartilage

Rats injected intra-articularly with fluorescent IGF-1 and dendrimer IGF-1 formulations were euthanized 2 and 6 days after injection. Image stacks from the intact medial femoral condyle taken 6 days after injection were reconstructed into three dimensions (Fig. 3E). IGF-1 fluorescent signal was visible throughout the width and depth of the femoral condyle for both dendrimer–IGF-1 formulations but was absent in the free IGF-1–treated samples. Two days after injection, IGF-1 was still present within the femoral cartilage (fig. S10). A related experiment was conducted ex vivo using 1-mm-thick bovine cartilage tissue disks (Fig. 4A), which more accurately represent human cartilage in structure and thickness (1 to 2 mm) (33). Explants were incubated for 2 or 6 days in dendrimer–IGF-1 (10 μM) before sectioning and visualization across the direction of diffusion using confocal microscopy (Fig. 4B). Area under the intensity profile indicated uptake of the formulation into the entirety of the cartilage, whereas distribution of intensity from right to left indicated diffusive penetration of the tissue. Even without any clearance mechanism in this experiment, free IGF-1 showed relatively little uptake and penetration of cartilage tissue relative to the dendrimer–IGF-1 formulations at both time points (Fig. 4C). After 2 days of uptake, the less charged Gen 4 35% PEG–IGF-1 formulation had diffused evenly throughout the tissue, whereas the more charged Gen 6 45% PEG–IGF-1 still exhibited a concentration gradient across the cartilage thickness. After 6 days of uptake, Gen 6 45% achieved a uniform concentration profile across the cartilage at nearly twice the average fluorescence intensity of Gen 4 35%. In summary, the more charged Gen 6 45% took 6 days to achieve a uniform distribution throughout the tissue, but by that time, a greater amount of total dendrimer formulation had been taken up from the bath than the Gen 4 35% formulation. The transport differences between the two formulations were more prominent in media lacking serum (fig. S11) in which electrostatic effects are inherently more pronounced.

Fig. 4 Dendrimer–IGF-1 conjugates exhibit full penetration of human-thickness cartilage within window of residence time in joint.

(A) Schematic of live bovine cartilage explant harvest and penetration assay. Fluorescent dendrimer distribution in cartilage was imaged after 2 or 6 days of incubation in Dulbecco’s modified Eagle’s medium (DMEM) + 10% fetal bovine serum (FBS) + cartilage media supplements. Red box indicates field of view (FOV). (B) Confocal microscope images of fluorescently labeled IGF-1 (purple) across the diffusion gradient (right to left, −y) of the tissue (shown in bright field). Arrow indicates direction of dendrimer transport. Scale bars, 200 μm. (C) Quantification of IGF-1 fluorescence intensity across the explant section. Average (Avg.) fluorescence intensity over the entire tissue section is shown. All images were taken under the same laser power, intensity, and offset.

Dendrimer–IGF-1 rescues cartilage from degeneration in a surgical rat model

On the basis of the results in Figs. 3 and 4, Gen 6 45% IGF-1 provided a higher dose of IGF-1 throughout the entirety of cartilage for a longer period of time than Gen 4 35% IGF-1, and thus, Gen 6 45% IGF-1 was selected for testing in a rodent OA model. Rats were injured in their right hindlimb knee by anterior cruciate ligament transection and partial medial meniscectomy (ACLT + MMx), a well-characterized and aggressive model of surgically induced OA (Fig. 5A) (2830). Two days after surgery, rats were injected into the affected knee with no treatment or free IGF-1 or Gen 6 45% PEG–IGF-1 [final joint concentration, 6 μM IGF-1 (~10 μM dendrimer)]. After 4 weeks, histological analysis based on the Osteoarthritis Research Society International (OARSI) histopathology initiative (Fig. 5B) (34, 35) was used to score joint degeneration in a blinded manner. Figure 5C shows representative images of histological sections of the medial knee used for scoring.

Fig. 5 Dendrimer–IGF-1 conjugate reduces cartilage degeneration 4 weeks after surgical traumatic joint injury.

(A) Schematic of a rat knee frontal section illustrating the ACLT + MMx surgery. Dashed box outlines the primary zone of lesion formation. (B) Schematic of surgery timeline and tissue processing procedures. I.A., intra-articular. (C) Representative toluidine blue/fast green–stained frontal sections of the medial femur and tibia. Area of degeneration outlined in red. Total and significant widths of degeneration are outlined in black and yellow, respectively. Matrix loss is shown as black arrowheads. MF, medial femur; MT, medial tibia; MM, medial meniscus; AC, articular cartilage; L, lesion. Scale bars, 500 μm. (D) Quantified area of degenerated cartilage tissue of the medial tibia for each rat, as a percentage of total cartilage area in the section. Degenerated tissue was defined as >50% cell death and loss of toluidine blue staining. (E) Width of cartilage degeneration at the joint surface (0% depth). (F) Width of cartilage degeneration at 50% cartilage depth. (G) Representative H&E-stained frontal sections of rat joints across different treatment, displaying regions of the lateral synovium characteristic of the given score (in parentheses). Cyan arrow indicates increased number of synovial lining cells, and black arrows indicate subsynovial proliferation. Scale bars, 100 μm. (H) Synovial inflammation scores (0 to 4) for each treatment. All data are means + 95% CIs, n = 7 to 9 rats, statistics by one-way ANOVA with Tukey HSD posttest for (D) and (E), and by Kruskal-Wallis test with Dunn posttest for (F) and (H).

Rats treated with Gen 6 45% PEG–IGF-1 after surgery showed a mean degeneration of 8.4% of the medial tibial cartilage area (Fig. 5D). This was significantly less (P = 0.017) than the untreated group, with a mean area degeneration of 23.7%. Free IGF-1 reduced area degeneration to 19.7%, but this was not significantly different from the untreated control. Similar trends were seen for widths of degeneration (Fig. 5E). At the cartilage surface (0% tissue depth), the mean width of degenerated tissue among rats treated with Gen 6 45% PEG–IGF-1 was 661 μm, compared to 1390 μm for IGF-1 alone and 1640 μm for untreated rats. Surface degeneration with Gen 6 45% PEG–IGF-1 treatment was statistically less than both other groups (P = 0.0023 versus untreated; P = 0.020 versus free IGF-1) and statistically equivalent (P = 0.57) to the sham operation, with a mean surface degeneration of 343 μm. Similar trends were seen for later-stage metrics of joint degeneration such as degeneration width at 50% tissue depth (Fig. 5F), but differences between treatments were not significant (P = 0.87 untreated versus Gen 6 45% PEG–IGF-1; P = 1.0 untreated versus free IGF-1). Data for these late-stage disease metrics appear to be bimodal, with a number of joints not exhibiting any deep degeneration or matrix loss across all treatment groups, suggesting that by 4 weeks after surgery, the disease had not yet progressed to those stages of damage in many animals.

Synovial inflammation in the lateral side of the joint capsule (Fig. 5G) was assessed by a blinded, board certified pathologist and scored on a semiquantitative 0 to 4 scale based on OARSI histopathology initiative guidelines for this model, taking into account number of synovial lining cell layers, subsynovial tissue proliferation, and inflammatory cell infiltrates (Fig. 5H). Synovial inflammation was relatively low overall at 4 weeks after surgery, although untreated and free IGF-1–treated rat joints had a significantly higher inflammation score than the sham surgery (P = 0.029 and 0.048, respectively). Dendrimer–IGF-1–treated joints had a mean score nearly 1 point lower than untreated joints (P = 0.22), 0.8 points lower than free IGF-1 (P = 0.33), and 0.3 points greater than the sham operation (P = 1.0).

Dendrimer–IGF-1 reduces osteophyte burden in a rat model of surgically induced OA

After animal euthanization, excised joints were imaged by microcomputed tomography (μCT). Osteophytes were identified and traced serially in two-dimensional (2D) frontal sections in a blinded manner following the methods of Batiste et al. (36, 37), based on protrusion from the normal bone contour and reduced bone mineral density. Figure 6A shows representative 2D μCT image sections with osteophytes identified by arrows, as well as 3D reconstructions of the images with the full volume of the osteophyte shaded in red. Total osteophyte volume was quantified for each rat (Fig. 6B). Rats that received surgery and no treatment showed high variability in osteophyte formation, with severe osteophytes in some rats and minor osteophytes in others. A single injection of free IGF-1 after surgery substantially reduced mean osteophyte burden by nearly 50% (1.10 to 0.60 mm3), but this effect was not statistically significant (P = 0.14). A single injection of Gen 6 45% dendrimer–IGF-1 was more effective, resulting in a mean total osteophyte burden of 0.23 mm3, which constituted a statistically significant reduction (P = 0.004) from untreated rats and was nearly equivalent to osteophyte burden of sham-operated rats (0.17 mm3; P = 0.99).

Fig. 6 Dendrimer–IGF-1 conjugate reduces osteophyte burden in surgically injured rats 4 weeks after surgery.

(A) Representative 2D (top) and 3D (bottom) μCT images showing osteophytes in red arrows or red shading. ROIs were serially drawn around osteophytes in sequential frontal image stacks and reconstructed into 3D to generate bottom images and measure osteophyte volume. (B) Total osteophyte volume in each joint across the four treatment conditions. Data are means + 95% CI, n = 7 to 10 rats as shown, statistics by one-way ANOVA with Tukey HSD posttest.


There is tremendous unmet medical need for a disease-modifying OA drug. Anabolic growth factors, such as bone morphogenetic protein 7, fibroblast growth factor 18 (FGF-18), and IGF-1, have shown potential for disease modification by decreasing chondrocyte loss and increasing matrix production in preclinical studies (12, 24, 3840). Anabolic disease-modifying strategies are particularly interesting in light of a recent finding by Heinemeier et al. (41) that revealed that cartilage renewal in adults is driven by aggrecan proteoglycan synthesis, with the underlying collagen matrix exhibiting no renewal in both healthy and osteoarthritic joints (41). Thus, the use of growth factors to provide enhanced production of aggrecan and protection of the chondrocytes that produce it seems a viable strategy for disease modification. Clinically, intra-articular FGF-18 increased cartilage thickness relative to placebo in a phase 2 OA trial (NCT01033994). However, the primary end point of the trial was not met (42). As free drugs, these growth factors and other biologics are hindered in efficacy by their short intra-articular half-lives and limited penetration of cartilage (9, 40, 43).

Here, in vivo imaging and penetration studies confirmed that the intra-articular pharmacokinetics of free IGF-1 are poor. With a joint half-life of just over 10 hours, even a relatively high 6 μM joint concentration will begin losing effect within 3 days. Moreover, IGF-1 could not penetrate more than ~20 μm into human-thickness tissue (~1000 μm), reaching only a small number of the chondrocytes requiring treatment for therapeutic gain.

Cartilage penetration is a crucial yet often neglected translational consideration in cartilage drug delivery. Any disease-modifying drug targeting chondrocytes must penetrate 1000 to 2000 μm of cartilage in humans to access all resident chondrocytes (33). However, many studies in the cartilage delivery field focus solely on mouse or rat cartilage, which is about 10 times thinner, between 50 and 200 μm depending on location in the joint (4446). As an avascular extracellular matrix, diffusion is the primary mode of transport available through cartilage, the rate of which can be enhanced by a factor of 2 by dynamic compression, such as that induced by joint motion (32). A given material will take a couple orders of magnitude greater time to diffuse through human tissue compared with rodent tissue based on tissue thickness alone. Thus, demonstrating penetration through thicker tissues, such as bovine cartilage, is absolutely necessary for translation of cartilage drug delivery technology intended to target chondrocytes (47). Some evidence of this discrepancy between rodents and larger mammals in our study can be observed by comparing data on the penetration of IGF-1 through rat and bovine cartilage. Free IGF-1 was observed within cartilage 2 days after injection into rat knees but was fully cleared from the tissue by 6 days. In 6 days, free IGF-1 could only penetrate ~20 μm through bovine cartilage, suggesting that if free IGF-1 was injected into a larger mammal such as a cow or human, then it would be cleared well before it could access more than ~1 to 2% of its target tissue.

Here, we presented a modified dendrimer nanocarrier capable of extending IGF-1 intra-articular half-life 10-fold and enabling full penetration of human thickness cartilage, thereby maintaining therapeutic amounts of the growth factor in cartilage for 30 days. We hypothesized that optimal drug delivery would improve IGF-1 efficacy and validated this hypothesis in a rat model of OA that has a high hurdle for detecting therapeutic cartilage protection (48). A single injection of free IGF-1 was unable to show statistically meaningful improvement in osteoarthritic rat knee joints. The Gen 6 45% PEG dendrimer–IGF-1 conjugate provided significant reduction in degenerated cartilage area, degenerated surface cartilage width, and total osteophyte volume relative to untreated rats, by factors of 3, 2.5, and 5, respectively. Moreover, Gen 6 45% PEG–IGF-1–treated rats were statistically equivalent to sham-operated rats by these metrics.

The partially PEGylated dendrimer–drug conjugate has several translational advantages over other cartilage drug delivery systems. PEG and PAMAM dendrimers are already manufactured at kilogram scales. PEG is a component in a wide variety of marketed pharmaceuticals. PAMAM and other cationic dendrimers have been used safely in clinical trials (NCT01577537, NCT03500627, and EUDRACT 2016-000877-19). The final PEGylated dendrimer–drug conjugate can be purified and characterized by standard chemical techniques, such as liquid chromatography, NMR, and MALDI-TOF mass spectrometry. This system is particularly suited for delivery in charged matrices with small mesh size, such as cartilage, due to both its adaptable charge and small size compared to other nanocarriers.

We have shown that the PEGylated dendrimer can incorporate biologic cargo without loss in bioactivity, making it ideal for delivery of biologics in the intra-articular space. The nondegradable bioconjugation scheme used is particularly advantageous for IGF-1 delivery, because it is designed to sterically hinder the N terminus of IGF-1, where the binding site for cartilage IGF-1 binding proteins is located. These IGF-1 binding proteins sequester IGF-1 from its target cellular receptors and are up-regulated with age and OA; they are thought to be responsible for a decrease in responsiveness to IGF-1 observed in aged or diseased rats and primates, including humans (4952). The bioconjugation chemistry used is flexible and could ostensibly be applied to nearly any molecule having a suitable synthetic handle. However, to incorporate a drug with an intracellular target, including nucleic acids and some small-molecule therapeutics, a biodegradable moiety should be introduced into the dendrimer-drug linker.

Tunable surface charge is another key feature of the partially PEGylated dendrimer–drug conjugate system. Previous work in our groups has shown that positively charged nanocarriers bind and penetrate cartilage efficiently via reversible electrostatic interactions with anionic cartilage matrix proteoglycans (11, 12, 15). Yet, the effects of degree of charge on these delivery properties have not been thoroughly explored. We systematically varied dendrimer surface charge by adjusting fractional end group PEGylation (0 to 60%) and observed marked effects on cartilage binding and cytotoxicity. Gen 6 45% PEG, which has three times as many free surface amines as Gen 4 35% PEG, exhibited greater total uptake into cartilage but slower tissue penetration, as expected on the basis of fundamental combined diffusion and binding transport kinetics. The more charged Gen 6 45% PEG formulation had a longer intra-articular half-life than Gen 4 35% PEG.

The partially PEGylated dendrimer system enables surface charge to be chemically tuned, characterized, and tested for biological performance. This feature enables the understanding of structure-property relationships in cartilage drug delivery. This platform approach provides the opportunity to optimize these systems for a range of intra-articular delivery goals based on desired therapeutic release profiles.

The highest-performing PEGylated dendrimer–IGF-1 conjugate that we identified in this study, Gen 6 45% PEG–IGF-1, showed no signs of toxicity at the cellular, tissue, or organ level. Unmodified PAMAM dendrimers have been shown to be toxic in a dose- and generation-dependent manner, which is consistent with our results (53). However, there are several reports of PEGylation at terminal amines mitigating or even nullifying this toxicity (18, 54, 55), which is consistent with our results. The negatively charged proteoglycans present in high quantities in cartilage and joint synovial fluid may provide some additional protection from cationic toxicity in this application. Local delivery by intra-articular injection enables the use of relatively small quantities of dendrimer and its associated drug, which mitigates concerns of systemic toxicity.

It is important to note the limitations of this study and additional questions that need to be investigated before clinical translation of this work. Although systemic biodistribution of these PEGylated dendrimers was not measured in this study, systemic bioelimination of 60% end group 3H-acetylated Gen 5 PAMAM dendrimers has been investigated by Nigavekar et al. (56). Within 7 days, 30% of total injected dose (ID) was excreted via urine, and 3% was excreted via feces. The PEGylated dendrimers in this study have a greater hydrodynamic radius; thus, excretion can be expected to shift from urine to feces. Nigavekar et al. (56) found that most organ uptake was localized to the liver, kidneys, and lungs at around 1 to 2% ID/g tissue at 7 days; these numbers decreased to 0.2 to 1.1% ID/g by 12 weeks after injection. We evaluated histological sections of these organs at 8 weeks after injection and found no signs of histotoxicity.

Although the rat ACLT + MMx is a well-characterized, rigorous model of OA (2830, 48), preclinical investigation in larger animals is necessary for regulatory approval. This is especially important given the aforementioned need to demonstrate penetration of thick cartilage in vivo.

In this study, animals were euthanized for histological evaluation of cartilage at 4 weeks, which coincided with both the duration of therapy provided by a single injection of Gen 6 45% PEG–IGF-1 and the reported onset of significant differences in tissue degeneration and osteophyte formation between operated and sham joints (29, 48). At this time point, we did not observe matrix loss or subchondral bone pathology in most of histological samples and so did not measure these features. Future animal studies should incorporate multiple injections over 8 to 12 weeks to observe the effect of this treatment on these late-stage end points.

In summary, we have identified a cationic nanoformulation capable of enhancing drug therapeutic lifetime to 30 days and cartilage penetration to at least 1 mm within articular joints. The dendrimer–IGF-1 formulation enhanced the efficacy of IGF-1 in protecting both cartilage and bone in a rat surgical model of OA, reducing the total area and width of medial tibial cartilage degeneration as well as total volume of osteophytes in the joint. The optimally PEGylated dendrimer–IGF-1 conjugates did not show any signs of toxicity on the cellular, tissue, organ, or organism levels. The pharmacokinetics of the nanoformulation are acceptable for repeat intra-articular injection (about monthly), and unlike most of the nanocarriers reported for OA, the formulation is capable of penetrating human-thickness cartilage. The chemistry used in dendrimer-drug conjugation is amenable to multiple classes of drugs, including small proteins, antibodies, small molecules, and nucleic acids. These results suggest the possibility of improving the efficacy of OA drugs using a partially PEGylated dendrimer–drug conjugate. This cartilage-penetrating nanocarrier could rejuvenate the field of OA drug development and accelerate the discovery of disease-modifying OA drugs.


Study design

Dendrimer–IGF-1 conjugates were evaluated for their pharmacokinetics in knee joints and efficacy in treating knee OA relative to free IGF-1 in a rat model. Surgical study outcomes used to evaluate therapeutic efficacy were prospectively designed to include both cartilage and bone pathology using histomorphometry and μCT imaging, respectively. We chose to evaluate joints 4 weeks after surgery to analyze both inflammatory response and early tissue pathology, according to characterization of OA development in this model (29, 30) and given that a single injection of dendrimer–IGF-1 remains active in the joint for around 4 weeks. Prospective power analysis was performed using G*Power analysis, and statistical analysis was performed using appropriate tests and multiple hypothesis testing corrections as indicated in the figures. Effect sizes were estimated on the basis of previous experiments in our group or published literature. Probability values of type I and type II error were set at 0.05 and 0.20, respectively. OA severity in this model is strongly affected by surgical skill, and so, subject order was blocked, not randomized, to account for this known variable. All histological and μCT measurement and scoring was conducted by a blinded investigator. Identifying animal numbers were converted to an unrelated system of letters by an individual uninvolved in scoring and were not revealed until all data analysis was complete. A total of six animals were removed from histomorphometric analysis because of widespread damage across the medial tibia, rendering quantitative measurement of cartilage impossible. A pathologist confirmed that this damage was mechanical in nature, likely because of severe joint instability generated by unintentional transection of multiple joint ligaments in the surgery. Joints were excluded while the investigators were blinded. Three of the excluded joints came from the untreated group, two of the excluded joints came from the Gen 6 45% PEG–IGF-1 group, and one of the excluded joints came from the free IGF-1 group.

Dendrimer–IGF-1 bioconjugation

PAMAM dendrimers, Gen 4 and Gen 6, were obtained from Dendritech through Sigma-Aldrich. PAMAM (350 nmol) was reacted with N-hydroxysuccinimide (NHS) methyl-PEG8 (mPEG8-SCM, Creative PEGWorks) and NHS-maleimide-PEG12 (SMPEG12, Thermo Fisher Scientific) in 0.1 M NaHCO3 buffer (pH 8.0). In a separate pot, Alexa Fluor 647 (AF647, Life Technologies) and N-succinimidyl S-acetylthioacetate (SATA; Thermo Fisher Scientific) in anhydrous dimethyl sulfoxide were added to IGF-1 (1 mg/ml; BioVision), each in a 2.5:1 molar ratio to produce IGF-1–SATA labeled with AF647. Both reactions were run for 2 hours at 25°C with rotation, protected from light. Maleimide-functionalized PEGylated dendrimers were purified by ultracentrifugation (Amicon Ultra, Millipore) using 0.1 M phosphate buffer (pH 6.8), and IGF-1–SATA was purified on a dextran desalting size-exclusion column (5 kDa; Thermo Fisher Scientific) with 0.1 M phosphate buffer as the eluent. The IGF-1–SATA was deacetylated to form IGF-1–SH by addition of 10% (v/v) 0.5 M hydroxylamine (Thermo Fisher Scientific) in 0.1 M phosphate buffer with 0.01 M EDTA (pH 8.0) for 50 min and then purified with a dextran size-exclusion column using 0.1 M phosphate buffer with 0.01 M EDTA (pH 6.8) as the eluent.

Fifty nanomoles of maleimide-functionalized PEGylated dendrimer was reacted with ~65 nmol of IGF-1–SH for 48 hours at 4°C with stirring. The reaction was then quenched by adding 2 μmol of cysteine for 2 hours. The mixture was purified by cationic exchange chromatography using a salt gradient of phosphate buffer at pH 6.8. About 60% of dendrimers were conjugated to IGF-1. IGF-1 yield based on starting quantity was about 25%.

Bovine cartilage explant harvest and culture

Young (1 to 2 weeks old) bovine knee joints were obtained from Research 87 (Hopkinton, MA), and cartilage was harvested from the trochlear groove and cultured by the methods in (24). Treatment conditions were blocked across explant location across the joint surface.

Cartilage explant penetration

A total of 750 μl of 10 μM near-infrared fluorescent (AF647) PEGylated dendrimer formulation (Gen 4 35% IGF-1, Gen 6 45% IGF-1, or free IGF-1) in DMEM (without phenol red), 10% (v/v) FBS, and 1% (v/v) penicillin-streptomycin was added to 6- by 1-mm cartilage explants. The explants were incubated for 48 or 144 hours at 37°C and 5% CO2 under gentle agitation. After incubation, cartilage explants were washed in PBS and sectioned to 80 μm using a vibrating microtome (Leica Biosystems). Sections were mounted on glass slides and immediately observed under a confocal fluorescence microscope (Nikon A1R scanning confocal microscope). Quantitative analysis was performed on maximum intensity projections of z-stack images taken from 80-μm sections.

Intra-articular pharmacokinetics

Twelve-week-old male Sprague-Dawley rats were anesthetized under isoflurane, shaved on both legs, and disinfected with three scrubbing routines of povidone iodine, followed by 70% (v/v) ethanol. Forty microliters of Gen 4 35% PEG–IGF-1, Gen 6 45% PEG–IGF-1, or free IGF-1, all matched in IGF-1–AF647 concentration (7.3 μM AF647 and ~20 μM IGF-1), in sterile isotonic saline was injected into both legs of the rat intra-articularly. An IVIS (Spectrum, PerkinElmer) and Living Image software (Caliper) were used to serially acquire and quantify fluorescence within each joint over a period of 28 days. Radiant efficiency data (units shown in Eq. 1) within a fixed anatomical ROI over time for each formulation type were fit to a single-phase exponential decay with a common plateau based on untreated animal background fluorescence.Embedded Image(1)

In a separate experiment under the same injection protocol, the distribution of fluorescently labeled drug or dendrimer-drug conjugate was observed in rat knees 2 and 6 days after injection. Joints were excised, split into tibia and femur, stained with a fluorescent antibody against aggrecan (Thermo Fisher Scientific), and imaged intact at the medial femoral condyle and medial tibial plateau in an Olympus FV1000 multiphoton confocal microscope throughout the depth of the cartilage tissue (120 to 160 μm) using an 840-nm Ti:sapphire laser and a 25× water objective with a numerical aperture of 1.25.

Rat surgical OA model: ACLT + MMx

All animal work was performed in accordance with protocols approved by the Institutional Animal Care and Use Committee at Massachusetts Institute of Technology (MIT) as well as the Animal Care and Use Review Office at the U.S. Army Medical Research and Materiel Command. Skeletally mature male Sprague-Dawley rats of 12 weeks of age (350 to 400 g; Taconic) were anesthetized by isoflurane, injected with subcutaneous meloxicam (1 mg/kg) and buprenorphine-sustained release (Bup-SR), and given eye ointment. The right hindlimb was shaved and disinfected with povidone iodine scrub, followed by 70% (v/v) ethanol in water. The rat was transferred to an operating table, rested on a heated recirculating water pad, and draped in a sterile manner. A shallow medial parapatellar incision was made with a #15 scalpel to expose the joint capsule, followed by a second incision to open the joint capsule. The incision was extended with iris scissors until the patella could be subluxated laterally, exposing the joint. Bleeding was mitigated by lavage with sterile saline and dabbing with sterile gauze. The ACL was exposed and identified by flexing the knee acutely and using surgical loupes. At this point, the sham operation was complete. For animals subjected to transection and meniscectomy, the ACL was transected with vannas microscissors, and transection was confirmed by the anterior drawer test. The medial meniscus was partially transected by inserting a #11 blade beneath the midpoint of the femoral condyle and pressing down gently on the tibia. After the completion of the procedure, the joints were thoroughly washed with sterile saline, and the patella was relocated. The joint capsule and skin were each closed separately with 5-0 polyglecaprone resorbable sutures. The skin sutures were reinforced with stainless steel wound clips and tissue glue. After surgery, rats were given warm subcutaneous saline (20 ml/kg), housed singly, allowed unrestricted movement, and put on an analgesic regimen of q24 subcutaneous meloxicam (1 mg/kg) and q48 Bup-SR (1 mg/kg) for 3 days.

Rats were divided into four groups: transection and no treatment (untreated), transection and free IGF-1 treatment (free IGF-1), transection and Gen 6 45% PEG–IGF-1 treatment (Gen 6 45%), and sham surgery (sham). Treatments were given by intra-articular injection into the affected joint 48 hours after surgery.


Processing of tissues for histology is presented in the Supplementary Materials and followed the methods of Schmitz et al. (34). The anterior and posterior tissues were embedded in the same cassette in paraffin and sectioned into a number of 5-μm frontal sections at three 250-μm steps within the joint. Sections were stained with toluidine blue/fast green, and the section containing most severe medial tibial lesion was identified by a blinded investigator for scoring. Gross pathological scoring followed the published guidelines of the OARSI Histopathology Initiative [Gerwin et al. (35)] and was performed quantitatively using ImageJ (National Institutes of Health) overlays on scanned digital slides (Leica Biosystems) by a blinded investigator (B.C.G.). Semiquantitative synovial inflammation grading also followed OARSI guidelines (35) and was performed by a blinded, board-certified pathologist (R.F.P.).

Microcomputed tomography

After fixation and before decalcification, excised rat joints were attached to the bore of a μCT scanner (GE eXplore CT 120) and scanned at 0.03° angles for a total of 1200 scans over a time period of 1.5 hours. Scans used an 80-kV x-ray potential, 32-mA current, and a 100-ms integration time. Image datasets were reconstructed into ROIs encompassing single joints using Microview software (Parallax Innovations).

3D positioning data were used to quantify total volume of osteophytes on each joint. 3D DICOM analysis software (OsiriX MD) was used by a blinded investigator (B.C.G.) to resection the images into a uniform frontal plane, perpendicular to an axis tangential to the femoral condyles and tibial plateau in the sagittal plane and an axis bisecting the femoral condyles and tibial plateau in the transverse plane. The blinded investigator then manually outlined osteophytes with ROIs in each contiguous frontal section. Osteophytes were defined as bone exhibiting both reduced bone mineral density and protrusion from the normal bone contour. The sections were then reconstructed to 3D, and total osteophyte ROI volume was calculated (36, 37).

Statistical analysis

For all experimentation shown, error bars indicates 95% CIs. Statistical analyses were performed with GraphPad Prism and SAS JMP. Two-sided P values less than 0.05 were considered statistically significant. Parametric data were analyzed by ANOVA methods, followed by Tukey HSD post hoc tests. Nonparametric data were analyzed by a Kruskal-Wallis test, followed by Dunn post hoc tests. Exponential decay constants from Fig. 3 were tested for statistical difference using the extra sum of squares F test, with a null hypothesis that data from all three treatments were adequately fit by a simple model with a common decay constant. Multiple hypothesis testing corrections are indicated in figure legends, if applied. Individual subject-level data for Fig. 2D and fig. S5 are shown in table S4.


Materials and Methods

Fig. S1. Representative cell cycle flow cytometry scatter plots for NIH/3T3 cells after 24-hour treatment with IGF-1 formulations.

Fig. S2. 1H NMR spectra of PEGylated dendrimer panel (without IGF-1).

Fig. S3. MALDI-TOF mass spectra for partially PEGylated dendrimer screen.

Fig. S4. Complete uptake data for partially PEGylated dendrimers into bovine cartilage explants.

Fig. S5. Complete CHON-001 cellular viability data for partially PEGylated dendrimers at a range of doses of dendrimers.

Fig. S6. Desorption of PEGylated dendrimers in solutions of different ionic strengths.

Fig. S7. Quantification of ex vivo bovine cartilage tissue viability staining in Fig. 2.

Fig. S8. Histology (H&E) of representative sections from likely contact organs for ~5- to 10-nm materials.

Fig. S9. Histology (H&E) of rats treated with dendrimers having suboptimal percentage of end group PEGylation.

Fig. S10. 3D reconstruction of multiphoton microscopy images of full-thickness cartilage from intact rat femurs harvested 2 days after injection of free IGF-1.

Fig. S11. Penetration of cartilage explants in DMEM (no protein in media).

Table S1. Characterization of percentage of PEGylation by NMR integral ratios.

Table S2. Characterization of percentage of PEGylation by mass addition according to MALDI-TOF data.

Table S3. Serum toxicology markers at 2 and 7 days after intra-articular injection of Gen 6 45% PEG–IGF-1.

Table S4. Individual subject-level data for Fig. 2D and fig. S5.


Acknowledgments: We acknowledge assistance from the Koch Institute Swanson Biotechnology Center, specifically the core facilities for Histology (Hope Tang facility), Animal Imaging and Preclinical Testing, Microscopy, Flow Cytometry, and Biopolymers and Proteomics. Special thanks go to K. Cormier, C. Condon, R. Bronson, and R. Porter for discussions on histological methods and J. Haupt and M. Jamiel for training on surgical techniques. Funding: This work was financially supported by DoD/PRMRP grant (W81XWH-14-1-0544). B.C.G. acknowledges a fellowship from the NSF. P.T.H. acknowledges the David H. Koch (1962) Chair Professorship in Engineering. Author contributions: B.C.G., A.J.G., and P.T.H. designed experiments and wrote the manuscript. B.C.G. conducted and analyzed experiments. S.W. contributed to animal studies and subsequent analysis. R.F.P. contributed to histopathological analysis. Competing interests: B.C.G. and P.T.H. are co-inventors on patent application PCT/US2018/039753 held/submitted by MIT that covers the partially PEGylated dendrimer–drug conjugates described herein. R.F.P. is a consultant to Medtronic, PLC on topics unrelated to the work described herein. P.T.H. is a cofounder of LayerBio Inc. and serves on the advisory board of Performance Indicator and Moderna Therapeutics. Data and materials availability: Any data associated with this study are available from the corresponding author upon reasonable request.

Stay Connected to Science Translational Medicine

Navigate This Article