In vivo multimodal imaging of hyaluronan-mediated inflammatory response in articular cartilage

Abstract

OBJECTIVE.

One driving factor in the progression to posttraumatic osteoarthritis (PTOA) is the perpetuation of the inflammatory response to injury into chronic inflammation. Molecular imaging offers many opportunities to complement the sensitivity of current imaging modalities with molecular specificity. The goal of this study was to develop and characterize agents to image hyaluronan (HA)-mediated inflammatory signaling.

DESIGN.

We developed optical (Cy5.5-P15-1) and magnetic resonance contrast agents (Gd-DOTA-P15-1) based in a hyaluronan-binding peptide (P15-1) that has shown anti-inflammatory effects on human chondrocytes, and validated them in vitro and in vivo in two animal models of PTOA.

RESULTS.

In vitro studies with a near infrared (NIR) Cy5.5-P15-1 imaging agent showed a fast and stable localization of Cy5.5-P15-1 on chondrocytes, but not in synovial cells. In vivo NIR showed significantly higher retention of imaging agent in PTOA knees between 12 and 72h (n=8, Cohen’s d>2 after 24h). NIR fluorescence accumulation correlated with histologic severity in cartilage and meniscus (ρ between 0.37 and 0.57, p<0.001). By using in vivo magnetic resonance imaging with a Gd-DOTA-P15-1 contrast agent in 12 rats, we detected a significant decrease of T1 on injured knees in all cartilage plates at 48h (−15%, 95%-confidence interval (CI)=[−18%,−11%] []) while no change was observed in the controls (−2%, 95%-CI=[−5%,+1%]).

CONCLUSIONS.

This study provides the first in vivo evidence that hyaluronan-related inflammatory response in cartilage after injury is a common finding. Beyond P15-1, we have demonstrated that molecular imaging can provide a versatile technology to investigate and phenotype PTOA pathogenesis, as well as study therapeutic interventions.

INTRODUCTION

The complex and poorly understood etiology of the onset of posttraumatic osteoarthritis (PTOA) continues to hinder the development of effective therapeutic interventions to prevent progressive joint destruction after injury.^1–7 Our ability to diagnose loss of tissue homeostasis will determine both how well we can phenotype PTOA and how effectively we can target therapeutic interventions.

Molecular imaging has potential to target with high specificity a pathway overtime and identify tissues in which the pathway is upregulated. In molecular imaging, we exploit the specificity of a molecule (antibody, peptide, small molecule) to create an imaging agent by tagging it with a molecule that can be seen with at least one imaging modality (e.g. MRI). In this way, we can see where the molecules accumulate in the tissue in vivo. While molecular imaging has been extensively used in tumor,⁸ rheumatoid arthritis⁹ and arteriosclerotic plaques,¹⁰ its application to OA has been anecdotal and its potential remains largely unexplored.^11–15

Here we show how molecular imaging can be used to target and characterize pro-inflammatory events for knee PTOA. In particular, we assessed a promising peptide, P15-1, a 15-mer peptide that functions by binding hyaluronan (HA) and modulating its role in inflammatory response.¹⁶ HA is an important glycosaminoglycan component of the pericellular matrix (PCM) of chondrocytes and many other types of cells.^17–20 In its normal high molecular weight (HMW) form (greater than 1 MDa),²¹ HA binds and clusters CD44 receptors to maintain homeostatic signaling and reduces pro-inflammatory and pro-fibrotic signaling by IL-1β, TNFα, TLR2, TLR4, and RHAMM.^22–24 In the presence of reactive oxygen and nitrogen species, HA also protects the cell membrane by acting as a sacrificial first target.²⁵ Increased HA synthesis is widely observed in tissues under inflammatory conditions,^{16, 26–29} and has been documented in cultured human articular chondrocytes subjected to IL-1β treatment.³⁰ This is believed to be a protective response to inflammatory stimuli, and therefore a possible biomarker of inflammation. Exogenous HMWHA is clinically used to relieve pain in osteoarthritis, and may have further therapeutic benefit based on the protective cell signaling effects of HMWHA.^{24, 31} The P15-1 peptide was identified in a phage display library as having HA-binding capability and a sequence with some homology to the HA-binding site of RHAMM.¹⁶ P15-1 can compete with RHAMM for HA binding.¹⁶ P15-1 peptide, especially when combined with HMWHA, shows anti-inflammatory effects on human chondrocytes cultured in inflammatory conditions (presence of IL-1β) and decreases catabolic signaling.³² In the present study, we employed labeled P15-1 to investigate its value in targeting elevated levels of HA associated with cartilage chondrocytes inflamed as a result of injury leading to PTOA.

The objectives of this work are: 1) Characterize in vitro imaging agents based on the P15-1 peptide for near infrared (NIR). 2) Validate the P15-1 imaging agents in vivo in two different animal models of PTOA for NIR and MRI.

METHOD

Contrast agents.

Four peptide conjugates of P15-1 were synthesized for our validation experiments, including the unlabeled parent peptide, a NIR dye-labelled version, a NIR dye-labelled scrambled version and a DOTA-conjugated version of the peptide (supplementary).

In vitro experiments:

Immortalized human articular chondrocytes (TC28a2, EMD Millipore) and sarcoma synovial cells (SW982, ATTC) were grown in monolayer cultures in complete medium.³² Cells were cultured in the presence or absence of recombinant human IL-1β (10 or 50 ng/ml in phosphate-buffered saline (PBS) and 0.1% bovine serum albumin (BSA) for 16 h). Then shifted to fresh medium and incubated with Cy5.5-P15-1 (2 μg/ml, 0.63 μM). In a first experiment, four independent sets of cells were incubated for 3, 6, 24 and 72 h after the addition of Cy5.5-P15-1 with each plate running triplicates. Cells were fixed and images acquired for each cell-type, condition and time. We measured cell viability with cell counting kit-8 (CCK-8, Dojingo) on TC28a2 and SW982 incubated for 72h with 10-fold the concentration used in vivo (Cy5.5-P15-1: 0.63 mM; Gd-DOTA-P15-1: 32.2 mM).

Fluorescence imaging analysis.

First, cell nuclei were segmented to identify individual cells. We calculated average Cy5.5-P15-1 signal in nuclei areas. For each cell-type, we used the distribution of cells incubated with vehicle after 3h of exposure to Cy5.5-P15-1 to set a threshold of signal intensity (third interquartile plus three times the interquartile range). Any cell whose NIR signal was above the threshold was classified as Cy5.5-P15-1 positive cell. Similarly, we identified a threshold for dead cells.

Animal models.

We induced PTOA in skeletally mature male Sprague-Dawley rats (14–16 weeks) by two different methods: transection of the anterior cruciate ligament (ACLT)³³ and ACL rupture by mechanical loading (ACLR) that mimics human injury.³⁴ ACLT animals underwent a sham operation in the contralateral limb; contralateral limbs of ACLR did not undergo any intervention. For imaging all contrast agent was administered intraarticularly (IA). All protocols involving the use of animals were approved by the New York University School of Medicine Institutional Animal Care and Use Committee.

Measurement of the elution curve of Cy5.5-P15-1.

Elution curve was measured with optical imaging using a Lumina XR-III scanner.³⁵ Eight rats underwent ACLT surgery and were housed for 8 weeks (Fig. 1). To reduce autofluorescence, animals were fed with alfalfa free diet (5V75, W.F. Fisher) and hair was removed from joints. Animals underwent bilateral IA injection of 50 μL of a 63 μM solution of Cy5.5-P15-1 in PBS. For in vivo imaging we used the lowest dose that provide good quality data (Supplementary). One animal was injected with the scrambled version (Cy5.5-ScP15-1, 50 μL of 63 μM solution). IA injection was performed using a Vevo 3100 ultrasound system (Fujifilm Visualsonics Inc.) following a protocol validated by our labs.³⁵ To provide reproducible positioning over time we used a custom designed device. Animals were scanned before IA injection and at 3, 6, 12, 24, 48 and 72 h after injection. We normalized the signal by the 3-h measurement. After 72 h, joints were harvested for optical/X-ray imaging and histology analysis.

Figure 1.

Animal grouping and study design. PTOA was induced either by transection of the anterior cruciate ligament (ACLT) or mechanical rupture of the ACL. Eight weeks after inducing PTOA, imaging agent was administered locally by IA injection in both injured and control knee. Animals were subjected to multiple imaging sessions and sacrificed at the end of the last imaging session (72 h for optical imaging studies and 48 h for MRI studies). Excised rat joints were flash-frozen, cryosectioned and subjected to histological analysis.

Optical images were analyzed by using Living Image 4.4 software (Caliper Life Sciences). A circular region of interest (ROI) of standardized size was drawn over the knee joint and average radiant efficiency was measured. To characterize the elution curves, we fitted the measured radiant efficiency to a three-compartment tracer-kinetic model (Supplementary). To identify optimal imaging times we estimated effect size (Cohen’s d).

Cryosectioning and histology analysis.

Seven excised rat joints (n=3 controls, n=4 ACLT) were flash-frozen and cryosectioned.³⁶ Sagittal sectioning was performed in three separate regions (lateral, central and medial). Cryosections were scanned using a fluorescence microscope to localize Cy5.5-P15-1 (IRD680 filter). We manually defined ROIs on cartilage and menisci to calculate mean signal on 5 to 7 slice per animal. Damage in cartilage and menisci were graded (cartilage: 0–5; menisci 0–3).³⁷ Synovitis was assessed on 6 control knees and 6 ACL-injured knees (n=3 ACLT, n=3 ACLR) from different animals using a validated grading system for synovial lining thickness with scores ranging 0–4 (healthy to severe synovitis).³⁸ The highest synovitis was used as joint-level outcome of synovitis. Sections from the same animals were stained for pro-inflammatory mediators with IL-6 (1:50, abcam) and iNOS (1:500, invitrogen) and the fraction of chondrocytes with positive staining calculated.

Ex vivo optical imaging.

To study the distribution of Cy5.5-P15-1 in tissue and analyze tissue specificity, we performed ex vivo imaging of optically cleared tissue incubated with Cy5.5-P15-1 on two limbs from two animals. Patellar, femoral condyle and posterior meniscus were isolated from a rat ACLT joint and cultured in complete medium;³² then incubated with Cy5.5-P15-1 (3.15 μM) for 15 min, washed and subjected to optical clearing.³⁹ Sequential z-stack images were acquired at 10× magnification.

Relaxivity of Gd-DOTA-P15-1.

We characterized relaxivity in solutions of Gd-DOTA-P15-1 at increasing concentrations (0, 32.2, 96.6, 193, 386 μM). The phantom was scanned with the same protocol used for animals. T₁-maps were obtained and a linear regression of relaxivity (1/T₁) and concentration was calculated.

In vivo validation experiments of Gd-DOTA-P15-1.

We validated Gd-DOTA-P15-1 in a set of 12 animals (n=6 ACLT; n=6 ACLR). Contralateral joints were used as internal controls. Animals were housed for 8 weeks. MRI were acquired within 2 days of IA injection of 50 μL of a 3.22 mM solution of Gd-DOTA-P15-1 and at 24 and 48 h after injection. We performed control-binding experiments. Three rats were injected with a combination of Gd-DOTA-P15-1 (3.22 mM) and Cy5.5-P15-1 (32 μM, a hundred-times lower) and underwent the same imaging protocol. Excised limbs underwent cryosectioning and fluorescence microscopy to detect Cy5.5-P15-1. Presence of Gd-DOTA-P15-1 was confirmed with scanning electron microscopy.⁴⁰ As negative controls for autofluorescence, we used the maximum intensity measured in four limbs of two animals where no Cy5.5-P15-1 was injected.

MRI protocol and image processing.

MRI was performed on a horizontal 7T Bruker system (Biospec 70/30, Bruker) equipped with a 700 mT/m gradient system using a 4-channel receive surface coil. MRI protocol included an inversion-recovery T1-weighted turbo-spin echo (TSE) sequence (TE/TR=12.5/5343 ms, turbo-spin acceleration factor=5, echo spacing=6.25 ms, 8 slices, slice thickness = 800 μm, slice gap = 400 μm, in-plane resolution=100×100 μm², inversion times (TI) = 0.15, 0.2, 0.3, 0.4, 0.5, 0.75, 1, 1.5, 2, 3, 4 and 5 s, acquisition time = 2:35 min/TI). All images were acquired in the sagittal plane oriented perpendicular to the line joining posterior femoral condyles.

Images for T1 calculation were registered using the cross-correlation of the Fourier transform of the images.⁴¹ Due to the varying contrast across TIs, image registration was performed for every two consecutive TIs. Finally, cartilage was segmented and average T₁ was calculated for each cartilage plate (patella, tibia and femur). Difference in T₁ values to the baseline acquisition, ΔT₁, were used to test accumulation of contrast agent.

Statistical analysis.

In vitro experiments were descriptive and used standard techniques of statistical inference to characterize mean, standard deviation and 95%-confidence intervals of the mean (bootstrapping, 1,000 repetitions).

To characterize elution curves, we fitted a three-compartment tracer-kinetic model to the average NIR signal measured in control and ACLT joints. We plotted residuals to ensure appropriate fit (normal probability and symmetry plots). We performed 1,000 repetition bootstrapping to characterize 95%-CI of the estimated parameters. To identify optimal imaging times, we performed an analysis stratified by time. Cohen’s d was calculated at each timepoint as the difference in measured signal between control and ACLT joints relative to their pooled standard deviation. A Cohen’s d over 2 was consider optimal for imaging.

To test if cartilage/meniscal damage was associated with Cy5.5-P15-1 accumulation we used a Generalized Estimating Equation (GEE) model (Signal~Histology grade). GEE model was selected to account for each animal providing multiple histological measures (exchangeable covariance matrix). We analyzed residuals per histology grade for bias. A separate analysis was performed for each cartilage plate and the meniscus.

We validated Gd-DOTA for in-vivo imaging. For each cartilage plate, group (ACL-injured vs. control), and timepoint we calculated the mean ΔT₁ and the 95%-confidence interval (bootstrapping). We used logistic regression to identify optimal threshold for ΔT₁ at 48 h differentiating ACL-injured knees and controls (ACLT~1+ΔT₁+animal). Logistic regression was validated using 5-fold cross-validation (1,000 realizations).

RESULTS

P15-1 imaging agents characteristics.

The fluorescently labeled conjugates (Cy5.5-P15-1 and Cy5.5-ScP15-1) exhibit high photostability in the NIR spectrum (excitation 684 nm, emission 710 nm) with a reported quantum yield of 0.2. For Gd-DOTA-P15-1, the relaxation rate (r1) was 2.4 mM⁻¹ s⁻¹ (Pearson’s r²=0.994, p<0.001). No toxic effects were detected for the imaging agents (see supplementary text and Table 2)

Table 2.

MRI cartilage T1 values. T1 values in articular cartilage¹

	Group	Pre-injection	24 h	48 h
	Control	1.11 (0.05)	1.13 (0.10)	1.10 (0.07)
	All ACL injury	1.19 (0.06)	1.08 (0.07)	0.98 (0.04) †
Femur	ACLT	1.19 (0.05)	1.07 (0.07)	0.95 (0.03) †
	ACLR	1.18 (0.03)	1.09 (0.04)	1.01 (0.08)
	Control	1.05 (0.05)	1.06 (0.08)	1.05 (0.03)
	All ACL injury	1.12 (0.07)	1.13 (0.06)	0.98 (0.06) †
Tibia	ACLT	1.14 (0.07)	1.15 (0.05)	0.97 (0.02) †
	ACLR	1.08 (0.04)	1.07 (0.02)	0.99 (0.05)
	Control	1.20 (0.10)	1.24 (0.13)	1.14 (0.10)
	All ACL injury	1.31 (0.08)	1.14 (0.12)	1.15 (0.09)
Patella	ACLT	1.33 (0.08)	1.18 (0.12)	1.16 (0.12)
	ACLR	1.29 (0.06)	1.05 (0.07)	1.15 (0.09)

¹T1 values in seconds represent the mean and standard deviation (in brackets) 95%-CI is shown in square brakets.

Bold face indicates significant differences with baseline examination.

^†Significant difference with 24h values (p<0.05, 2-sided unpaired ttest).

In vitro experiments.

Figure 2 represents the fraction of chondrocyte and synovial cells that showed Cy5.5-P15-1 accumulation. TC28a2 cells showed fractions of Cy5.5-P15-1 positive cells that increased both over time and by treatment with IL-1β. In comparison to TC28a2 cells, a lower fraction of SW982 did accumulate Cy5.5-P15-1. Fraction of dead cells was 1–5% for both cell-types under all conditions. Distinctive patterns of accumulation of Cy5.5-P15-1 on human chondrocytes were identified (Fig. S1). TC28a2 and SW982 had over 90% viability after 72 h incubation with high dose of Cy5.5-P15-1 or Gd-DOTA-P15-1 (Supplementary).

Figure 2.

In vitro cell localization of Cy5.5-P15-1. Summary of chondrocyte and synovial cell fraction showing increased Cy5.5-P15-1 signal after treatments with IL-1β (0, 10 and 50 ng/ml). Measurements were taken at 0, 3, 6, 24 and 72 h and each includes at least 2,000 cells. Threshold of Cy-5.5-P15-1 signal to consider a cell positive was defined based on cells with no IL-1β at 3 h. Bars represent 95%-confidence interval of the mean over three independent measures. Inserts represent detail of the early changes in time.

Characterization of elution curves in vivo with NIR imaging.

Elution curve of Cy5.5-P15-1 from both limbs of 8 animals over a period of 72 h is shown in Figs. 3A–C; including data from the scrambled version (Cy5.5-ScP15-1) in an ACLT limb. Optimal parameters of the tracer-kinetic model fit to the data (solid lines in Fig. 3C) are summarized in Table 1. Longest half-life of imaging agent (log(2)/r-) in the joint was 11.3 h (95%-CI [10.2h–12.4h]) and 12.4 h (95%-CI [10.8h–13.6h]) for Cy5.5-P15-1/sham and Cy5.5-P15-1/ACLT groups. Retention fraction was over 12% (95%-CI [6%–21%]) for the ACLT group and below 0.25% (95%-CI [0.1%–3.4%]) for the control group. Cohen’s d between Cy5.5-P15-1/ACLT and control groups (Cy5.5-P15-1/sham, Cy5.5-ScP15-1/ACLT) increased over time reaching effect-size larger than two 24h after injection (Fig. 3D).

Figure 3.

In vivo NIR imaging of Cy5.5-P15-1 elution from ACLT and sham hindlimbs. A, in vivo microCT scan showing rat positioning by using a custom designed device that holds the knee in flexed with 20 degrees of internal rotation allowing for a frontal view of the tibiofemoral space. B, serial fluorescent images of a representative animal over time including excised joints (fluorescent and X-rays) after last in vivo imaging point. C, elution curves for Cy5.5-P15-1 on ACLT (n= 8, red) and sham joints (n= 8, blue) and Cy5.5-ScP15-1 on an ACLT joint (n= 1, orange). Error bars represent the 95%-CI of the mean signal. Measured values were fitted to a three-compartment tracer-kinetic model (solid lines (Supplementary). D, Plot of the effect size (Cohen’s d, d_s) of Cy5.5-P15-1 signal intensity between ACLT and control groups (Cy5.5-P15-1/sham, Cy5.5-ScP15-1/ACLT). Bars represent the 95%-CI of d_s calculated by bootstrapping with 1,000 repetitions. Solid line represents a linear fit to d_s; dashed line indicates an effect-size of 2. Equation escribes d_s from two samples of mean (X₁ and X₂), sizes n₁ and n₂ and standard deviation SD₁ and SD₂

Table 1.

Tracer kinetic model of elution curves¹

	r+ (h−1)	r− (h−1)	f+	f0
Cy5.5-P15-1/ACLT	0.59 [0.47–0.63]	0.056 [0.051–0.064]	0.68 [0.65–0.72]	0.003 [0.001–0.034]
Cy5.5-P15-1/sham + ScCy5.5-P15-1/ACLT	0.53 [0.42–0.60]	0.061 [0.056–0.068]	0.44 [0.39–0.48]	0.116 [0.062–0.211]

¹Values are represented with 95%-CI calculated using bootstrapping (1,000 realizations)

Histology analysis.

Cy5.5-P15-1 localized on chondrocytes, predominantly in those situated in the deep cartilage layer and calcified cartilage; and in the extracellular matrix of superficial cartilage (Fig. 4A–C, 5). Cy5.5-P15-1 also accumulated in the posterior meniscus, mostly in fibrochondrocytes of the inner zone and, to a lesser extent, in fat pad and ligament insertions (Fig. 5).

Figure 4.

Histological analysis of Cy5.5-P15-1 biodistribution in joint tissues. A-C, fluorescence microscopy showing Cy5.5-P15-1 localization on representative slides of the different grades of damaged cartilage observed. A, Control left knee joint from an animal that underwent ACLT in the right knee joint that shows no cartilage damage (OARSI score 0) and little accumulation of Cy5.5-P15-1; B, ACLT knee joint of the same animal (right knee) with OARSI score 3 in the tibial cartilage and 2 in the femoral cartilage; and C, ACLT knee joint with OARSI score 2 for both femoral and tibial cartilage. D, box plot of the average NIR signal measured in cartilage and menisci for different histology grades of damage. Individual measurements are indicated by circles. Asterisk indicates statistically significance compare to the score zero group (p 0.05).

Figure 5.

Histological details of Cy5.5-P15-1 biodistribution in ACLT joint tissues. A, 40× detail of the tibial cartilage showing localized imaging agent in articular cartilage (AC) and subchondral bone (SB). Cy5.5-P15-1 accumulated predominantly on the radial zone. B, Left. 63× detail of chondrocytes in the radial zone of tibial AC showing Cy5.5-P15-1 localization. Right, same cells (in red) overlapped over the collagen extracellular matrix (white) captured by auto fluorescence. C, 63× detail showing imaging agent on chondrocytes. D, 10x detail of posterior meniscus (PM) showing localized Cy5.5-P15-1 around areas of damage (F=Femur, T=Tibia). E, Image shows accumulation of Cy5.5-P15-1 in the meniscal insertion. F-G, Cy5.5-P15-1 was found in the insertion of the anterior cruciate ligament (ACL) in ACLT rat knee joints. F was acquired at 10× and G at 40×.

8 weeks after ACLT, cartilage showed moderate damage (score 1 to 3). Cy5.5-P15-1 accumulated in cartilage of ACLT joints compared to controls (overall +40.9%, 95%-CI [36.5%–55.4%]). Uptake of Cy5.5-P15-1 varied for different cartilage plates, with femur showing larger accumulation (+57.0%, 95%-CI [35.1%–80.0%]) than tibia (+16.6%, 95%-CI [11.4%–21.8%]). Cy5.5-P15-1 uptake increased with damage in both cartilage and menisci (Fig. 4D). GEE models showed that histology grade was a predictor of Cy5.5-P15-1 signal intensity (Table 3). Significant Spearman’s correlation with histology scores was found for femoral (ρ=0.57) and tibial cartilage (ρ=0.32) and meniscus (ρ=0.48).

Table 3.

Summary of GEE models for Cy5.5-P15-1 and histology scores

		Estimate	Standard error	95%-CI
Femur	Intercept	132.0	7.6	[117.1, 146.9]
	Histology score	53.5	12.3	[29.4, 77.6]
	Correlation (ρ)	0.15±0.17
Tibia	Intercept	146.8	8.0	[131.1, 162.5]
	Histology score	23.9	6.3	[11.6, 36.2]
	Correlation (ρ)	0.12±0.11
Meniscus	Intercept	247.2	12.9	[221.0, 273.3]
	Histology score	41.1	7.7	[25.4, 56.7]
	Correlation (ρ)	0.17±0.19

Synovitis scores were elevated in ACL-injured knees (mean score=1.83, 95%-CI [1.33–2.33]) compared to controls (mean score=0.33, 95%-CI [0–0.67]). ACL-injured joints had markedly large fraction of chondrocytes labeled with iNOS (40.5–61.7%), and IL-6 (43.4–54.7%) compared to controls (iNOS: 3.1–4.7%, IL6: 0.3–1.3%).

Ex vivo penetration of Cy5.5-P15-1.

Ex vivo imaging of optical-cleared samples incubated with Cy5.5-P15-1 showed tissue penetration in Figure S3. 15 min incubation with Cy5.5-P15-1 before the clearing process led to uniform penetration for the whole depth of cartilage and meniscus.

Validation of Gd-DOTA-P15-1 as MR contrast agent in PTOA.

Image registration resulted in an average displacement of less than a voxel (mean 85 μm, range 0 to 254 μm). 15 out of 72 T₁-measurements presented through-plane motion and were not included. At least eight measurements were available for each time point for ACL injured and controls.

Average T₁ values are summarized in Table 2. Over time, cartilage T₁ decreased in the ACL injured but not in controls (Fig. 5, Table 2). At 48 hours, T₁ dropped in ACL injured knees −12.5% (95%-CI=[−17.0%, −8.0%]) in the tibia, −17.7% (95%-CI=[−20.1%, 15.2%]) in the femur and −15.1% globally (95%-CI=[−18.3%, −11.8%]). T₁ did not change in control groups (−2.1% globally, 95%-CI=[−4.6%, +0.5%]). Both injury models (ACLT and ACLR) showed the same trend of T₁ decrease. At 48 h, logistic regression models (ACL injury ~ constant+ΔT₁+animal, constant=28.9, 95%-CI [12.8, 45.1]; ΔT₁ coefficient −28.7 s⁻¹, 95%-CI=[−53.6, −3.7]; Animal coefficient 0.10, 95%-CI=[−0.22, 0.41]) identified an optimal threshold for ΔT₁ of −0.092 s to differentiate between ACL injured and controls (~10% of baseline T₁). Five-fold cross-validation resulted on an average ΔT₁ threshold of −0.091 s (95%-CI [−0.77 to −0.115]) with an accuracy of classification in left-out data of 94.5%. No control limbs had a ΔT₁<−0.09 s, while T₁ decreased more than 0.1 s for all but one ACL-injured limbs, whose baseline T₁ value (1.08 s) was in the T₁ range in controls. Interestingly, at 48h all but one ACL-deficient animal included had T₁ values in femur bellow 1 s, while all controls had T₁ values above 1 s.

In control-labeling experiment, we were unable to detect Cy5.5-P15-1 in limbs 48 h after injection (Fig 5C). Electron microscopy showed similar biodistribution of Gd-DOTA-P15-1 than what we found for Cy5.5-P15-1.

DISCUSSION

In this work we show that molecular imaging can provide a powerful method to study in vivo activation of a specific pathway in PTOA. To our knowledge, this study provides the first direct in vivo evidence for cartilage after joint injury.

Assessment of inflammation in cartilage in vivo is challenging because of its thin anatomy (~200 μm in rats). For our approach, we used a combination of MRI and NIR optical imaging. NIR imaging provides very sensitive, quick and inexpensive assessment of imaging agent distribution in vivo, although with ~1 mm resolution due to light scattering. Fluorescent probes can be also detected in histology when using the cryo-tape sectioning that does not require tissue fixation or decalcification. MRI can assess tissues with resolutions of 50–100 μm in vivo with excellent soft tissue contrast to enhance thin cartilage plates. However, MRI is less sensitive to detect contrast agents (>100 μM) than NIR optical imaging (>50 nM).

A second obstacle in measuring inflammation in cartilage is its avascularity, which causes standard MRI methods to diagnose inflammation to fail in cartilage. The standard MRI method for inflammation uses gadolinium imaging agents to identify increased stromal vascularity in inflamed tissues. Contrast agents can only reach cartilage by diffusion through synovium or subchondral plate, which is an inefficient mechanism of delivery. As an example, delayed gadolinium enhanced MRI of cartilage (dGEMRIC) requires a double dose of contrast agent, exercise and waiting times of 90 mins between contrast agent administration and MRI. In dGEMRIC, T₁ maps are used as an indirect measure of glycosaminoglycan distribution in cartilage.

In vitro experiments have shown these imaging agents interact strongly with the target. In our time-lapse experiments, Cy5.5-P15-1 remained localized for at least 72 h with little change in its distribution or intensity. Interestingly, expression of targets with affinity for Cy5.5-P15-1 is a fast process with half-lives of around one hour, and once expressed shows little signs of motility. Given the high collision rate between a Cy5.5-P15-1 molecule and a cell target (~700 collisions/s, supplementary), our data indicates that localization of P15-1 is likely mediated through a target that become available in a short period of time. Our experiments also show that P15-1 has a greater affinity for chondrocyte than synovial cells, which was also confirmed in our histology analysis. P15-1 binding to cells was increased by treatment with Il-1β. Finally, ex vivo experiments showed that Cy5.5-P15-1 efficiently penetrates the extracellular matrix; fifteen minutes incubation was enough to provide uniform distribution across the extracellular matrix of cartilage and meniscus. Thus, it is reasonable to assume that Cy5.5-P15-1 penetrated all tissues accumulating only where targets were available. P15-1 compounds have likely low toxicity as shown by our results on cell viability and previous toxicity studies of P15-1 in rabbits (Patent US-10,449,229-B2). Furthermore, we found a clear association between tissue damage and Cy5.5-P15-1 accumulation for both femoral cartilage and meniscus, indicating the relationship between structural damage and pro-inflammatory environment. The relatively large fraction of pro-inflammatory state of chondrocytes shown by iNOS and IL-6 immunohistochemistry is likely driving the accumulation of P15-1-derived contrast agents in cartilage.

Contrast agent Gd-DOTA-P15-1 allowed us to detect pro-inflammatory response in vivo in cartilage. T₁ values pre-contrast agent showed a significant difference between ACL-deficient and control knees. Native T₁ is modulated by the macromolecular environment in cartilage and has been reported to change with cartilage degradation.^42–47 However, differences in pre-contrast T₁ are not linked to inflammation. On the contrary, post-contrast changes in T₁ are only due to accumulation of contrast agent, with changes in relaxivity (1/T₁) being proportional to its concentration. Interestingly, we found that T1 measured at 48 h has potential to identify concentration of contrast agent. All animals that had a ΔT1≤−0.1 s (optimal threshold from logistic regression) had also an average T1 at 48h below 1s, while all animals with ΔT1>−0.1 showed average T1 over 1s. If confirmed, it will become possible to assess use a single T1 measure 48 h after injection to detect pro-inflammatory events in cartilage.

There is increasing evidence that inflammatory mechanisms play a central role in mediating the development and progression of PTOA after injury.⁴⁸ HMWHA is considered to be a suppressor of inflammation,⁴⁹ and thus maintaining HA integrity in the extracellular matrix may attenuate inflammatory and catabolic events and help restore tissue homeostasis in cartilage. P15-1 binds HA and competes with HA binding to RHAMM, a HA receptor that has been shown to induce inflammation and fibrosis in different tissues.¹⁶ P15-1 suppressed catabolic and inflammatory events in cultured human articular chondrocytes treated with IL-1β, even under conditions of low RHAMM expression, by reducing activation of the p38 MAPK signaling pathway.³² We showed that IL-1β increases expression of HAS-1, HAS-2 and Hyal-2 indicating a pro-inflammatory response of chondrocytes.³⁰ It has been described that p38 MAPK plays a major role in stimulating inflammatory and catabolic events.⁵⁰ We have proposed that P15-1 binding to HA at sites of active pro-inflammatory signaling may contribute to the cell-protective properties of HMWHA, leading to greater CD44 clustering and the prevention of factors binding and signaling through inflammatory receptors, such as TLRs; and stabilizing the PCM of chondrocytes.³² Remarkably, our imaging studies confirm that increased accumulation of P15-1 can be found in vivo on articular chondrocytes of the inner layer of femoral cartilage, following injury to the joint.

There are only a few studies that have used molecular imaging to measure inflammation in OA, all targeting specific receptors expressed on activated macrophages.^{12, 13} Yang et al. used a positron emission tomography (PET) tracer based on a synthetic peptide with high affinity for formyl peptide receptor 1 (FPR1) to detect inflammation in an animal model of OA (monoiodoacetate). The authors demonstrated the capability and specificity of the tracer to bind to macrophages in vitro; however, there was not a direct correlation of PET signals to macrophages in vivo.¹² On the other hand, Kraus et al. used etarfolatide (EC20) nuclear imaging (SPECT/CT) that targets the folate receptor to detect activated macrophages in human knee OA joints; and found a clear correlation between etarfolatide quantities and both radiographic and symptom severity of knee OA.¹³ Etharfolatide imaging has helped to identify six additional synovial markers associated with macrophage activation.⁵¹ Our study, however, targets a different aspect of inflammation, providing unique technology to assess pro-inflammatory environment in cartilage that cannot be imaged with any of the existing imaging agents.

There were several limitations of this study. For one, it represented a small sample size. By design, this study was intended to be an initial examination of the potential utility of a HA-binding peptide (P15-1) as a contrast agent to image pro-inflammatory events in knee OA. The study focuses on a single time point of the disease and further studies are required to evaluate early stages of PTOA. MRI is the image modality that offers a good compromise between resolution and sensitivity to contrast agent. However, a resolution of 100 μm can, despite careful segmentation, led to partial volume effect. We used two different imaging agents for optical and MRI imaging. However, control-binding experiments have shown that high concentration of Gd-DOTA-P15 saturate targets for Cy5.5-P15-1, confirming that both imaging versions mark the same targets. Finally, although we have shown an IL-1β dose-dependent P15-1 binding to chondrocytes in vitro we cannot exclude that binding could also be caused by other factors such as growth factors that might upregulate hyaluronan and synthesizing enzymes. However, it is very unlikely that growth factors would be the primary source of activation of P15-1 binding in the pro-inflammatory environment of the injured joint. Several studies reporting proteomic analysis of synovial fluid after joint injury showed elevated presence of inflammatory markers but no significant increase in growth factors compared to controls.⁵² However, additional studies are warranted to determine the specificity of IL-1β for mediating the binding of P15-1 to chondrocytes.

In summary, this study is the first in vivo indication for the involvement of pro-inflammatory response in cartilage after injury. Our results provide insights into the pathogenesis and role of pro-inflammatory signaling in OA, showing that hyaluronan-mediated inflammatory response is a common finding. Beyond the P15-1 peptide, we have demonstrated that molecular imaging methods can provide a versatile technology to investigate and phenotype OA pathogenesis, as well as to study therapeutic interventions.

Figure 6.

In vivo MR imaging of Gd-DOTA-P15-1 on ACL injured and control hindlimbs A, Representative T₁ maps on an ACLT and control knee over time showing T₁ decrease on ACLT hind limb and no decrease on control. B, box plots representing the decrease in T₁ in all cartilage plates in the ACL injured but not in controls. Circles represent individual measurements. C, left, representative ACLT fluorescent cryosections from animals injected with a combination of Gd-DOTA-P15-1 and Cy5.5-P15-1 in a 100:1 ratio or only Cy5.5-P15-1 after 72h; right, representative scan electron microscopy showing gadolinium distribution on cartilage. SEM of cryosections was performed without any coating and using a backscatter detector in a Zeiss Gemini FESEM microscope.