In vivo imaging of protease activity in arthritis: A novel approach for monitoring treatment response




Sensitive noninvasive strategies for monitoring treatment response in rheumatoid arthritis (RA) would be valuable for facilitating appropriate therapy and dosing, evaluating clinical outcome, and developing more effective drugs. Because different proteases are highly up-regulated in RA and contribute significantly to joint destruction, in the present study we investigated whether such enzymes are suitable in vivo imaging biomarkers for early evaluation of treatment response in a murine model of RA.


Using a protease-activated near-infrared fluorescence (NIRF) imaging “smart” probe, we examined the presence and distribution of fluorescence in arthritic joints of mice with collagen-induced arthritis by both noninvasive fluorescence imaging and histology. Proteases that target the Lys–Lys cleavage site, including cathepsin B, activate probe fluorescence. Treatment monitoring data were obtained following methotrexate (MTX) therapy.


Twenty-four hours after intravenous injection of the protease sensor, affected toes and paws of arthritic mice showed significantly higher fluorescence intensity than did toes and paws of healthy mice. Fluorescence from the protease probe and cathepsin B antibody histologic staining were localized in the vast majority of cells in the inflamed synovium. In arthritic animals treated with MTX (35 mg of MTX/kg 48 hours prior to probe injection), a significantly lower fluorescent signal (inflamed paws 50%, inflamed toes 70%) was observed as compared with untreated arthritic animals.


Protease-activated NIRF probes are sensitive means of imaging the presence of target enzymes in arthritic joints and can be used for early monitoring of treatment response to antirheumatic drugs such as MTX.

Significant advances in the treatment of rheumatoid arthritis (RA) have been achieved in the past decades, with the increased spectrum of available therapies resulting in improved treatment outcome. However, individual responses to treatment vary. Thus, selecting and optimizing treatment for individual RA patients is challenging (1, 2). One potential approach is to monitor early treatment response. Unfortunately, this is currently limited. Measurement of serum parameters of inflammation, which may not accurately reflect disease activity within joints (3, 4). Serial radiographic assessment of bone demineralization and joint space narrowing to monitor treatment response typically begins late in the course of the disease, limiting its usefulness as a method for short-term assessment of treatment effects (5, 6). Other imaging techniques, such as magnetic resonance imaging and ultrasound, may have potential roles in the evaluation of the disease in the future (3, 4), but again typically provide information on anatomic changes that are later manifestations of RA.

Matrix-degrading enzymes, produced primarily by synovial fibroblasts and synovial macrophages, have a substantial role in the destruction of arthritic joints. These proteolytic enzymes contribute significantly to the invasion and degradation of articular structures (7–9). Several enzymes of different protease families, such as matrix metalloproteinases (MMPs) (10, 11) and cysteine proteases, e.g., cathepsin B, are involved in this process (12–14). Since matrix-degrading enzymes are highly up-regulated in RA (7–14) and effective therapies must decrease the destructive activity of these proteases, we hypothesized that these enzymes are suitable targets for in vivo imaging of early treatment response. The ability to measure such indirect biomarkers of disease activity would be useful in both the preclinical and clinical realms of RA treatment.

Previously, we developed near-infrared fluorescence (NIRF) imaging “smart” probes, which markedly increase their level of fluorescence after target interaction (15). These selective protease probes are biocompatible, and their fluorescence is quenched when administered intravenously until activated by enzymatic cleavage in the target tissue. This allows localization of signal in areas of activity with low fluorescence background in adjacent tissues that do not activate the probes and, importantly, increased signal indirectly correlates with increased proteolytic activity. Imaging probes based on this concept, with specificities for Lys–Lys bonds (such as cathepsins B, L, and S, and plasmin) and other probes with selectivity for specific proteases, such as cathepsin D or MMP-2, have been developed and characterized. They have previously been applied to imaging of enzyme activity in tumors, atherosclerosis, and inflammation, demonstrating their substantial value for noninvasive imaging of molecular markers of disease (15–18). Such molecular reporters may serve as indicators for objective assessment of treatment response, for stage and severity of disease, and prognosis, may facilitate early diagnosis, and may allow for more effective drug development.

The goals of this study were to examine whether methotrexate (MTX) treatment would alter protease activity, to determine whether protease-activated NIRF imaging probes are sensitive reporters for imaging the presence of target enzymes in arthritic joints, and to determine whether such probes could be used as molecular reporters for early monitoring of treatment response in RA. For this purpose, untreated and MTX-treated mice with collagen-induced arthritis (CIA) were injected with protease-activated imaging probes. Proteases that target the Lys–Lys cleavage site, including cathepsin B, activate probe fluorescence. Thereafter, fluorescence reflectance imaging was performed at different time points after administration, the fluorescence intensity of inflamed and healthy paws was analyzed, and fluorescence activation was examined histologically.


Protease-activated probe.

The imaging probe was synthesized as previously described (15). Briefly, multiple self-quenching fluorochromes (Cy5.5; excitation and emission maxima 675 nm and 694 nm, respectively) were bound to a long, circulating graft copolymer consisting of a poly-L-lysine backbone sterically shielded through multiple methoxypolyethylene glycol (MPEG) side chains (Figure 1A). Cleavage of the probe by enzymes such as cathepsin B and other proteases that cleave the Lys–Lys bond released Cy5.5, resulting in fluorescence activation that could be detected in vivo (15).

Figure 1.

A, Schematic diagram of the near-infrared fluorescence (NIRF) imaging probe. Multiple Cy5.5 fluorochromes were bound to a graft copolymer backbone of poly-L-lysine, to which methoxypolyethylene glycol (MPEG) side chains were attached. Due to interactions between the fluorochromes, fluorescence quenching occurred. Enzymatic cleavage of the backbone by proteases such as cathepsin B released Cy5.5 and resulted in activation of fluorescence. B, Raw NIRF image of a mouse with collagen-induced arthritis in the right fore paw, obtained 24 hours after probe injection, showing high fluorescence intensity in the affected extremity. C, Color-coded NIRF image of B, superimposed on white light image. A well of Cy5.5 dye (16 nmoles/ml), seen above the right hind paw, was used for standardization.

Animals and arthritis model.

All animal studies were approved by the institutional animal care committee. Thirty-six male DBA/1J mice, 4–5 weeks of age, were obtained from The Jackson Laboratory (Bar Harbor, ME). CIA was induced according to the procedure described by Wooley (19): bovine type II collagen (Sigma, St. Louis, MO) was dissolved in 0.1M acetic acid at a concentration of 2 mg/ml by stirring for ∼3 hours at 4°C. To prepare the adjuvant, 3 mg heat-inactivated Mycobacterium tuberculosis H37Ra (Difco, Detroit, MI) was added to 1 ml Freund's incomplete adjuvant (Difco). Equal volumes of the collagen solution and the adjuvant were emulsified at 4°C. Thirty mice were immunized by 3 intradermal injections of the emulsion at the base of the tail (50 μl each). Three weeks after the first immunization, the procedure was repeated. Within 2 weeks after the second immunization, ∼70% of the mice developed arthritis, characterized by erythema and swelling of the paws. In the majority of mice, the disease progressed to 2 or more paws.

MTX treatment and administration of the protease-activated probe.

To study the effect of MTX on protease activity in arthritic mice, 20 DBA/1J mice with CIA were distributed randomly into 2 groups. One group received intravenous injections of saline (arthritis, untreated; n = 10). Another 10 mice were treated with MTX (Ben Venue Laboratories, Bedford, OH) (intravenous injection of 35 mg/kg body weight). A third group of nonimmunized mice was injected intravenously with saline (no arthritis, untreated; n = 6). Forty-eight hours after treatment, the mice received a protease-activated probe (which cleaves Lys–Lys bonds) intravenously. The probe has a molecular weight of 480 kd (15), and the parent compound, MPEG–poly-L-lysine, has a blood half-life of 20 hours in humans.

In vivo optical imaging and image analysis.

Fluorescence reflectance imaging was performed using a previously described system (20). Briefly, the imaging system consisted of a light-tight chamber equipped with a 150W halogen white light source, a 615–645–nm excitation band-pass filter (Omega Optical, Brattleboro, VT), and light diffusers to distribute the photons over the field of view. A 12-bit monochrome charge-coupled device camera (IS440; Kodak, Rochester, NY) equipped with a zoom lens and an emission band-pass filter at 680–720 nm (Omega Optical) was used for image acquisition. For imaging, the animals were anesthetized with ketamine/xylazine (80 mg/12 kg body weight). The animals were imaged in the prone position together with a reference standard that contained 16 nmoles/ml free NIRF dye (Amersham Biosciences, Piscataway, NJ). Between imaging sessions, the mice were kept on a heating pad until they fully recovered from anesthesia. A template was used to facilitate reproducible positioning of the mice and the standard. Imaging was performed 10 minutes, 3 hours, 6 hours, 12 hours, and 24 hours after injection of the protease-activated probe, which contained a total of 2 nmoles of quenched NIRF dye. This dose was based on doses that had previously been used in tumor model systems. Signal intensity of this class of activated probes is linear with respect to enzyme concentration (16). White light images were taken first (image acquisition time 0.075 seconds), followed by fluorescence detection (image acquisition time 120 seconds). Images were analyzed with commercially available software (Digital Science 1D; Kodak) for image field correction, and by Image J software (NIH, Bethesda, MD) for region of interest (ROI) analysis. ROIs were placed over the paws, the skin (snout), and the reference standard. The fluorescent signal of the mouse paws, expressed as relative fluorescence intensity (RFI), was calibrated to the standard, subtracting the skin as background. Custom software (Center for Molecular Imaging Research, Charlestown, MA) was used to generate color-coded images. Immediately after the last imaging session, animals were killed, paw weights were obtained, and the samples were prepared for histologic study.

Histologic analysis.

To examine cathepsin B activity histologically as a contributor to protease-based probe activation, the inflamed paws of 3 arthritic mice that received an intravenous injection of saline (control) and of 5 arthritic mice that received an intravenous injection of protease-activated probe (2 nmoles Cy 5.5) were prepared 24 hours after injection, snap-frozen in TissueTek embedding medium (Sakura, Zoeterwoude, The Netherlands), and stored at −70°C. The tissue samples were cut into 8-μm–thick slices using a CM 1900 cryotome (Leica Microsystems, Wetzlar, Germany). Serial sections of each tissue sample were collected on SuperfrostPlus-precoated microscope slides (Menzel-Glaeser, Braunschweig, Germany) and were first analyzed for NIRF with an inverted epifluorescence microscope (Axiovert; Zeiss, Thornwood, NY) equipped with a cooled, charge-coupled camera (SenSys; Photometrics, Tucson, AZ). For fluorescence excitation, a 660–680–nm band-pass filter was used. Fluorescence was inspected with a 700-nm long-pass filter (Omega Optical). The sections were then stained with primary antibodies against cathepsin B (rabbit anti-rat; US Biological, Swampscott, MA) by the alkaline phosphatase–anti–alkaline phosphatase (APAAP) method, as previously described (21).

Briefly, the sections were fixed in acetone, blocked for endogenous peroxidase activity using H2O2, covered with a fish gelatin, glycine, and normal goat serum in Tris buffer (pH 7.6) to block nonspecific binding, washed in Tris buffer, and incubated for 45 minutes at room temperature with the primary antibody or an isotype control diluted 1:50 in Tris buffer. The slides were rinsed in Tris buffer and incubated for 30 minutes at room temperature with a secondary goat anti-rabbit IgG antibody (Jackson ImmunoResearch, Hamburg, Germany) diluted 1:400 in Tris buffer. Incubation with the APAAP complex (Dako, Glostrup, Denmark) for 30 minutes at room temperature (diluted 1:100 in Tris buffer) was followed by the addition of substrate using the New Fuchsin substrate system (Dako). Color development was stopped, after examining the slides microscopically, by adding a mixture of 10 mM Tris buffer and 1 mM EDTA. The slides were mounted immediately for microscopic analysis in comparison with serial hematoxylin and eosin (H&E)–stained sections.

Statistical analysis.

Data are expressed as the mean ± SEM. Statistical significance was measured by unpaired t-test. P values less than 0.05 were considered significant.


At several fixed times after intravenous injection of the imaging probe, optical-reflectance NIRF imaging was performed. Figure 1B shows a raw NIRF image and Figure 1C shows a color-coded NIRF image (obtained 24 hours after probe injection) superimposed on a white light image of a mouse with arthritis that affected the right fore paw. In contrast to nonarthritic paws, arthritic paws showed high fluorescence intensity, which indicated intense enzymatic cleavage of the imaging probe in inflamed paws, which results in fluorescence activation.

To investigate fluorescence activation on the histologic level, tissue sections of arthritic and nonarthritic paws were examined for NIRF and in parallel by immunohistochemistry using antibodies specific for cathepsin B. Figure 2 shows images of serial cryostat sections of inflamed, hyperplastic synovium from a mouse with CIA 24 hours after injection of the imaging probe. Fluorescence of the probe, as seen on the NIRF image (Figure 2A), was detected in the hyperplastic synovium of inflamed joints. Most notably, the vast majority of synovial cells showed fluorescence. In the hyperplastic synovium of uninjected control mice (control for autofluorescence), no NIRF signal was found (results not shown). Serial section staining (Figure 2B) revealed that virtually all cells expressed cathepsin B protein. This helped confirm that the fluorescent signal (Figure 2A) was secondary to protease expression from cells. In Figures 2C and D, serial synovial sections stained with H&E and an isotype control of the anti–cathepsin B antibody as a negative control are shown. The isotype control, using the same staining procedure with the exception of a nonreactive isotype of the primary antibody, showed only minimal nonspecific staining. As additional negative and positive controls, we also stained mouse liver tissue, known to express cathepsin B. Using a nonspecific isotype of the primary antibody, the liver sections showed negligible nonspecific staining, whereas using the anti–cathepsin B antibody as the primary antibody resulted in intense staining throughout the liver tissue (results not shown).

Figure 2.

Serial cryosections of inflamed, hyperplastic synovium from a mouse with collagen-induced arthritis. A, Near-infrared fluorescence (NIRF) image of activated protease probe. B, Anti–cathepsin B antibody stain for immunoreactive protease. C, Hematoxylin and eosin stain. D, Isotype control of antibody stain. The large structure in the upper right is bone. Almost all the cells in the hyperplastic synovium exhibited cathepsin B activity, as shown in B. Images were obtained 24 hours after injection of the NIRF probe.

To examine whether MTX treatment altered activation of fluorescence, fluorescence reflectance imaging of untreated and MTX-treated arthritic mice was performed at different times after intravenous injection of the imaging probe. Figure 3 shows white light, raw NIRF, and color-coded NIRF images superimposed on the white light images of untreated mice (Figures 3A and C) and MTX-treated mice (Figures 3B and D) 24 hours after injection of the protease-activated probe. Fluorescence intensities of arthritic toes and paws of MTX-treated mice were found to be substantially lower than those of untreated mice.

Figure 3.

White light, raw near-infrared fluorescence (NIRF), and color-coded NIRF images superimposed on white light images of mice with collagen-induced arthritis (CIA) 24 hours after injection of the probe. A, Untreated mouse with arthritis in 3 toes of the right fore paw (arrowheads). B, Methotrexate (MTX)–treated mouse with CIA in a toe of the right fore paw (arrowhead). C, Untreated mouse with arthritis in the right hind paw. D, MTX-treated mouse with CIA in the right hind paw. MTX reduced fluorescence intensity from arthritic toes and paws compared with untreated mice.

Figure 4 shows the time course of fluorescence intensity, presented as RFI based on ROI analysis, in arthritic and nonarthritic toes (Figure 4A) and paws (Figure 4B) of untreated and treated mice after probe injection. The RFI of nonarthritic toes and paws increased only slightly with time. In contrast, the NIRF signal of arthritic toes and paws of untreated mice increased rapidly due to intense cleavage of the protease-activated probe. After 24 hours, inflamed toes and paws of untreated arthritic mice showed a 5-fold and 7-fold higher RFI than the toes and paws, respectively, of untreated healthy mice (P < 0.001). Compared with the fluorescence intensity in untreated arthritic mice, the signal measured in arthritic toes and paws of MTX-treated animals was found to be significantly lower at 12 hours and later (P < 0.001); after 6 hours, the level of significance was P < 0.001 for toes and P = 0.01 for paws. After 24 hours, inflamed toes and paws of treated arthritic mice showed a 3-fold and 2-fold lower RFI than inflamed toes and paws, respectively, of untreated arthritic mice.

Figure 4.

Time course of fluorescence intensities in arthritic and nonarthritic toes and paws of mice after probe injection. A, Fluorescence from arthritic toes of untreated and methotrexate (MTX)–treated mice and healthy toes of untreated mice. B, Fluorescence from arthritic paws of untreated and MTX-treated mice and healthy paws of untreated mice. Values are the mean ± SEM relative fluorescence intensity (RFI). MTX treatment significantly decreased fluorescence at 6, 12, and 24 hours postinjection (P < 0.001 for toes and P = 0.01 for paws at 6 hours; P < 0.001 for toes and paws at 12 and 24 hours). At 24 hours, the RFI in the toes and paws of untreated arthritic mice was increased 5-fold and 7-fold, respectively, compared with healthy mice.

Measurement of paw weights in all arthritic groups revealed no significant differences between treated and untreated groups at the time of imaging (P = 0.97 and P = 0.3 for fore paws and hind paws, respectively, between treated and untreated groups). There was no correlation between relative fluorescent signal intensity and paw weight (r values ranged between 0.01 and 0.1 for the various groups).


The noninvasive visualization and quantitation of biologic processes on the cellular and molecular levels for use as biomarkers of disease activity is an emerging approach to therapy monitoring. In this context, the results of the present study demonstrate that protease-activated NIRF imaging “smart” probes can serve as sensitive reporters for imaging treatment response to antirheumatic drugs such as MTX. An advantage of activated probes in contradistinction to targeted probes, e.g., fluorescence-labeled antibodies with a specificity for a certain target, is that the background signal is much lower due to the very low fluorescence in the activated probe's native state, while the high fluorescence postactivation at the target site provides additional molecular fidelity of the signal. In addition, another form of signal amplification occurs using “smart” probes because one target enzyme molecule can activate multiple signal molecules (15).

The imaging technologies used in this study are inexpensive and have high throughput capability (22). The properties of near-infrared light propagation allow for the direct recording of signal after penetration through 1 cm of tissue, making human peripheral joints easily accessible via surface reflectance techniques. Additionally, 3-dimensional localization and quantification of near-infrared fluorochromes at nanomolar concentrations, after penetration of up to 10 cm of tissue, is feasible by tomographic techniques that use advanced reconstruction (23). Thus, no technologic hurdles exist that would prevent the implementation of near-infrared recording of “smart” probe activation in patients with RA.

A protease-activated imaging probe was chosen because matrix-degrading enzymes, including cysteine proteases such as cathepsin B, are highly up-regulated in patients with RA, contributing significantly to the destruction of arthritic joints (7–14). Besides directly degrading major joint components, cysteine proteases are capable of activating MMPs. Thus, cysteine proteases also contribute to a cooperative process among the proteolytic enzymes (for review, see ref. 24). In addition, expression of cathepsin B appears to be restricted predominantly to the synovial cells attached to cartilage and bone at sites of joint erosion in RA patients (12, 13). Furthermore, cysteine proteases are expressed in the synovium of RA patients at a very early stage of the disease and in established erosive RA, and are absent from normal synovium (8, 25, 26). The protein expression of cathepsin B and procathepsin L correlates with the score of chronic inflammatory activity (14). Moreover, enzymatic activity of cysteine proteases in the joints of rats with antigen-induced arthritis is positively correlated with the severity of joint destruction as well as with inflammation parameters (27), and parallels disease activity (28). Thus, cysteine proteases are suitable targets, especially for the early monitoring of treatment response.

The imaging probe used in this work cleaves Lys–Lys bonds. Cathepsin B is one protease that cleaves such bonds, and we have demonstrated here that in the CIA model there is colocalization of fluorescence imaging signal and cathepsin B expression at the histologic level. However, additional proteases, such as cathepsins L and S and plasmin, cleave the Lys–Lys bond and may contribute, in part, to signal activation in this and other disease applications. Alternate probe constructs may allow increased specificity for cathepsin B cleavage, similar to increased specificity shown for MMP-2 (16). Importantly, the probe used here more accurately reflects total available protease load in vivo than probes with alternate specificities, and may thus be a more useful clinical biomarker of disease status and response to therapy, given the heterogeneity of gene expression in the clinical setting compared with animal model systems.

MTX was selected as a proof-of-concept antirheumatic drug to evaluate the suitability of enzyme-activated NIRF imaging probes for early monitoring of treatment response. MTX is probably the most commonly used drug in the treatment of RA and serves as a gold standard (1, 2). Furthermore, studies demonstrate the ability of MTX, like other treatments, to decrease the activity of proteolytic enzymes in the synovium (29, 30). Using the smart probe paradigm, we have been able to show that MTX significantly decreases fluorescence intensity of inflamed toes and paws of mice with CIA up to 2–3-fold 24 hours after probe injection. This decrease could easily be determined by measuring changes in fluorescence activity (Figures 3 and 4). Fluorescence intensity measured at the target site in this work reflects a combination of probe delivery and activation. However, under the current experimental conditions, the dominant effect is most likely due to changes in enzyme activity, and is unlikely to reflect changes in probe delivery since paw weight measurements showed no significant differences between treated and untreated arthritic paws. Additionally, there was no correlation between fluorescent signal intensity and paw weight in all groups, also suggesting that there were no differences in leak (contributing to edema) between treated and untreated animals during the course of the experiment. The MTX dose administered to the mice exceeded the clinical dose. However, comparably high MTX doses are needed to induce a sufficient treatment response in this animal model (31).

In conclusion, noninvasive optical imaging using enzyme-activated NIRF probes such as ones that are selective for the proteases that cleave Lys–Lys bonds, e.g., cathepsin B, may be useful for evaluating the presence of target enzymes in arthritic joints and for the early monitoring of treatment response to antirheumatic drugs such as MTX. Potential clinical benefits include the rapid titration of therapy choice, therapy dosing, quantitative early evaluation of clinical outcome, and more effective drug development.


The authors would like to thank Lee Josephson and Alexei Bogdanov, Jr., for critical review of the manuscript. We also gratefully acknowledge the excellent technical assistance of Wibke Ballhorn of the Division of Rheumatology and Clinical Immunology, Department of Internal Medicine I, University of Regensburg, Regensburg, Germany.