Introduction
Rheumatoid arthritis (RA) is a chronic destructive inflammatory disease of the joints. Although the disease pathogenesis remains unclear, there is significant evidence implicating T cells and B cells in the early initiating steps of disease and innate immunity in its chronic, slow progression [
1]. Both genetic and environmental factors contribute to the development of RA [
2], and the disease shows a steady progression of synovial hyperplasia and neovascularization, mixed mononuclear and granulocytic cellular infiltration, damage to articular cartilage, bone remodeling, and proliferation of both synovial and extraarticular fibroblasts [
1,
3]. This manifests clinically as swelling, erythema, and pain, and can progress to decreased bone density and obvious joint architecture changes.
Of current importance in the development of anti-arthritic drugs is the ability to discriminate between disease-modifying anti-rheumatic drugs (DMARDs), which affect arthritis pathogenesis and progression, and non-DMARDs, which may show palliative effects and symptom relief in the absence of affecting disease progression. DMARD treatments include antiproliferative drugs (for example, leflunamide and methotrexate) or cytotoxic drugs (azathioprine) as well as agents that interfere with TNFα, such as anti-TNF biologics (adalumimab, etanercept, infliximab). Inhibitors of p38 mitogen-activated protein kinase (MAPK) have also been shown to reduce TNF levels and affect disease pathogenesis in animal models of RA [
4‐
8], with some more modest effects in patients showing DMARD efficacy [
4,
9] limited by dose-dependent toxicity. Whereas p38 MAPK inhibitors significantly decrease underlying inflammation and bone destruction, cyclooxygenase-2 (COX-2) inhibitors, such as celecoxib, and other nonsteroidal anti-inflammatory drugs (NSAIDs) are better at providing symptom relief than at altering disease progression [
10,
11].
A variety of rodent arthritis models have been used to study arthritis disease progression and the impact of promising new therapies [
12]. These models include the current gold standard approaches using type II collagen-induced arthritis in both the mouse and rat, and have been used extensively for benchmarking novel therapies while being routinely validated against current standards of care (methotrexate and prednisolone). Their utility is limited, however, as mouse collagen-induced arthritis models require specific disease-susceptible inbred mouse strains (that is, DBA/1 and B10.RIII) in order to develop arthritis, placing a heavy emphasis on the early inductive phase of disease. In contrast, newer models that bypass the cognate immunity step in disease induction by using inducing antibodies to trigger chronic disease, such as the collagen antibody-induced arthritis (CAIA) model, provide a more straightforward and rapid means of producing disease pathology that is both independent of the mouse strain and can be used with transgenic or knockout mice [
13‐
16]. Although the mouse CAIA model does not have the extensive history and therapeutic validation of the collagen-induced arthritis model, there is growing support for the relevance of autoantibodies in mouse arthritis [
17‐
23] and in human arthritis [
24‐
27], and there is particular evidence suggesting the importance of autoantibodies at disease onset [
28].
Regardless of the particular rodent model used to study disease mechanisms, current non-invasive standard readouts of disease severity - such as paw thickness/volume or clinical score grades - do not provide a quantitative biological readout of the cellular/tissue-specific processes contributing to disease progression. For instance, paw thickness uses dimensional changes in the paw as a surrogate marker for underlying edema and inflammation, while clinical score assessment is a subjective assessment of paw swelling and erythema. Although these readouts can be useful measures of disease severity, they emphasize the edema component of disease rather than the underlying synovial proliferation, inflammatory cell infiltration and, osteoclast-mediated bone resorption. Paw swelling or clinical scores therefore do not discriminate well between DMARD and non-DMARD treatments such as NSAIDs. For instance, the non-DMARD anti-inflammatory COX-2 inhibitors (a type of NSAID) routinely demonstrate efficacy in a variety of rodent arthritis models, as determined by paw swelling/clinical score [
5‐
7,
12,
29,
30]. Because of this, there is significant reliance upon (terminal) histopathology to discriminate DMARD activity from NSAID activity when assessing new drugs.
In the present article, we build upon recent advances in optical tomographic imaging and near-infrared (NIR) agents [
31‐
36] to test the hypotheses that biological imaging of molecular optical biomarkers of inflammation and bone turnover would provide superior non-invasive (nonterminal), quantitative readouts for underlying disease pathology, and that - when used in combination with optical tomographic imaging - the CAIA model should provide robust and quick discrimination between DMARD and non-DMARD treatments.
Our studies illustrate the ability of three-dimensional fluorescence molecular tomographic (FMT) quantification to discriminate between DMARD and non-DMARD effects. For instance, neither clinical score, paw thickness, nor multiple plasma biomarkers could differentiate between a p38 MAPK inhibitor and the COX-2 inhibitor celecoxib, while FMT quantification using NIR agents to detect cathepsin, matrix metalloprotease (MMP), or bone resorption activity yielded a clear discrimination between these two classes of treatment. FMT results agreed well with histopathologic scoring of inflammation, and both FMT and histology measures identified clear deficiencies in clinical score and paw-swelling assessments of disease. Optical tomographic imaging of disease biology offers a non-invasive, nonterminal measure of disease that strongly correlates with the underlying pathology of the CAIA model and allows for discriminating between DMARD and non-DMARD therapeutics.
Materials and methods
Experimental animals
Specific pathogen-free female BALB/c mice (4 to 6 weeks of age, 18 to 20 g) were obtained from Charles River (Wilmington, MA, USA) and were housed in a controlled environment (72°F; 12 h:12 h light-dark cycle) under specific-pathogen free conditions with water and food provided ad libitum. All experiments were performed in accordance with VisEn IACUC guidelines for ethical animal care and use.
Therapeutic studies with the collagen antibody-induced arthritis animal model
BALB/c mice were injected intravenously with 4 mg arthrogen-collagen-induced arthritis monoclonal antibody cocktail (Clones D1, F10, A2 and D8 to collagen type II; Chemicon, Temecula, CA, USA), according to the manufacturer's instructions. Measurable morphological changes were determined by paw thickness measurement using a digital Vernier caliper (VWR, West Chester, PA, USA) on days 4, 6, and 8. Observational clinical scores (scale from 0 to 3) were also made based upon the following criteria of redness and swelling: 0 = no swelling or redness (normal paws), 1 = swelling and/or redness in one digit or in the ankle, 2 = swelling and/or redness in one or two digits and ankle, and 3 = entire paw is swollen or red.
Beginning on day 3 post antibody cocktail injection (prior to signs of disease), cohorts of CAIA mice (n = 12 per group) were treated daily (8 or 15 days) with either prednisolone (10 mg/kg per oral, twice daily), a p38 MAPK inhibitor (SD0006; 15 mg/kg per oral, twice daily), and celecoxib (15 mg/kg per oral, twice daily). Two additional groups, healthy mice (n = 12) and arthritic mice (n = 12), received vehicle treatment only (0.5% aqueous methyl cellulose and 0.025% Tween-80) and served as controls.
Fluorescent agents for the detection of inflammation
Three commercially available imaging agents (VisEn Medical Inc., Bedford, MA, USA) were used to measure disease and therapeutic efficacy in CAIA. For assessing the inflammatory infiltrate, two NIR protease-activatable agents were used, one activated by cathepsins (ProSense750) and the other activated by a family of MMPs (MMPSense680), including MMP-3, MMP-9, and MMP-13. These agents were administered via intravenous route (2 nmol (fluorophore) in 150 μl saline) in all imaging studies. A third NIR imaging agent that detects changes in bone associated with disease (OsteoSense680) was used to image and quantify bone loss. For MMPSense680 and OsteoSense680, the 2 nmol dose of fluorophore corresponds to 2 nmol substrate or pamidronate, respectively. For ProSense750, the 2 nmol dose of fluorophore corresponds to ~0.1 nmol substrate.
Imaging arthritis disease progression
CAIA and control mice were injected intravenously with ProSense750 or MMPSense680 on day 7 following injection of collagen antibody cocktail. OsteoSense680 was injected in additional studies on both day 7 and day 14. At the time of imaging (24 h post agent injection), mice were anesthetized using an intraperitoneal injection of ketamine (100 mg/kg) and xylazine (20 mg/kg). CAIA and control mice were then imaged with the FMT 2500™ fluorescence tomography
in vivo imaging system (VisEn Medical) using fluorescence tomographic scanning capabilities as described previously [
37]. Briefly, the anesthetized mice were carefully positioned in a prone position in the imaging cassette. Both hind paws were elevated on a resin block (designed to mimic optical scattering and absorption properties of the mouse's body) to allow larger tomographic scanning fields for simultaneous imaging of both paws. A NIR laser diode transilluminated the hindpaws, with signal detection occurring via a thermoelectrically cooled charge-coupled device camera placed on the opposite side of the imaged animal. Appropriate optical filters allowed collection both of fluorescence and excitation datasets, the entire imaging acquisition requiring 4 to 5 minutes per mouse.
Fluorescence molecular tomographic reconstruction and analysis
The collected fluorescence data were reconstructed by FMT 2500 system software (TruQuant™; VisEn Medical) for the quantification of the fluorescence signal within the paws. Three-dimensional regions of interest were drawn to encompass each foot and subregions of the foot. A threshold was applied identically to all animals equal to twice the mean paw fluorescence (nanomolar) of the control, nonarthritic mice to minimize low-intensity, background fluorescence. The total amount of ankle, midfoot, toes or total paw fluorescence (in picomoles) was automatically calculated relative to internal standards generated with known concentrations of appropriate NIR dyes. For visualization and analysis purposes, the FMT 2500 system software provided three-dimensional images and tomographic slices.
Histopathology
The right ankle from each animal was fixed in 10% neutral buffered formalin for 24 hours at 20°C, followed by decalcification in Immunocal™ (Decal Chemical Corporation, Tallman, NY, USA) for 7 days at 20°C. Decalcified joints were then paraffin embedded, sectioned twice (4 μm each), and stained with H & E for general evaluation or toluidine blue for specific assessment of cartilage changes. The ankles were evaluated via histopathology and scored for inflammation, cartilage damage, pannus and bone resorption according to previously published criteria [
38].
For inflammation, scores were as follows: 0 = normal, 1 = minimal infiltration of inflammatory cells in the synovial and/or periarticular tissues, 2 = mild infiltration with mild edema, 3 = moderate infiltration (including joint space) with moderate edema, 4 = marked infiltration with marked edema, and 5 = severe infiltration with severe edema.
For cartilage damage, scores were as follows: 0 = normal, 1 = loss of toluidine blue staining with no chondrocyte degeneration/loss and/or matrix disruption, 2 = loss of toluidine blue staining with minimal chondrocyte degeneration/loss and/or mild matrix disruption in some affected joints, 3 = loss of toluidine blue staining with moderate chondrocyte loss and obvious (depth to deep zone) matrix loss in affected joints, 4 = loss of toluidine blue staining with marked (depth to tide mark) chondrocyte and matrix loss, and 5 = loss of toluidine blue staining with severe (depth to subchondral bone) chondrocyte loss and matrix loss in affected joints.
For bone resorption, scores were as follows: 0 = normal, 1 = minimal (small areas of resorption in the medullary trabecular or cortical bone, not readily apparent on low magnification, and rare osteoclasts), 2 = mild (increasing areas of resorption in medullary trabecular or cortical bone, not readily apparent on low magnification, with osteoclasts more numerous), 3 = moderate (obvious resorption of the medullary trabecular and cortical bone, without full-thickness defects, lesion apparent on low magnification, and osteoclasts more numerous), 4 = marked (full-thickness defects in the cortical bone, marked loss of medullary trabecular bone, numerous osteoclasts), and 5 = severe (full-thickness defects in the cortical bone, severe loss of medullary trabecular bone).
Immunoassay analysis of plasma biomarkers
Plasma MMP-3, a soluble marker for joint pathology, was quantified by the R&D System (Minneapolis, MN, USA) Quantikine mouse MMP-3 (total) Immunoassay (catalog number MMP300) according to the manufacturer's instructions. Plasma cytokines and chemokines - eotaxin, granulocyte colony-stimulating factor (G-CSF), granulocyte-macrophage colony-stimulating factor, GRO/KC, IFNγ, leptin, IL-1α, IL-1β, IL-2, IL-4, IL-5, IL-6, IL-9, IL-10, IL-12p70, IL-13, IL-17, IL-18, IP-10, MCP-1, MIP-1β, RANTES, TNFα and vascular endothelial growth factor - were assessed using a multiplex Luminex-based assay from Millipore (catalog number MPXMCYTO-70K-PMX24; Billerica, MA USA) with the addition of 1× Complete® protease inhibitor cocktail (catalog number 11697498001; Roche, Indianapolis, IN, USA).
Statistical analysis
Data are presented as the mean ± standard error of the mean. Significance analysis of in vivo paw fluorescence was conducted using a two-tailed unpaired Student t test when two groups were analyzed or a one-tailed analysis of variance Scheffe multiple-comparison post test. P < 0.05 was considered significant.
Discussion
Various rodent models of inflammatory arthritis are used in research and drug development, due to their similar pathology and/or pathogenesis to human RA, their ease of use and reproducibility, and their ability to predict drug efficacy in humans. Although rodent models share a number of important morphologic and immunologic features with RA, they progress rapidly and are heavily reliant upon the acute inflammatory response. Despite this, rodent arthritis models have contributed greatly to the overall knowledge of RA and have led to important advances in therapeutic intervention. Yet disease assessment in rodent models relies heavily upon subjective endpoints (clinical scoring and paw swelling) that emphasize the edema component of disease and capture little or none of the molecular processes that drive differential cellular infiltration and/or bone resorption. Given the emergence of new targeted molecular therapeutic agents (reviewed in [
42,
43]), improved methods for reliably and objectively detecting and quantifying disease and therapeutic responses without sacrificing the animal (real time) are warranted.
Proteases play a central role in the human RA disease process. We therefore reasoned that protease-activatable NIR imaging agents [
44,
45] could serve as a sensitive means for reporting disease initiation and therapeutic responses. Such agents have been used to detect protease upregulation in a number of disease states, including cancer [
46,
47], asthma [
37,
48], atherosclerosis [
32,
49], and inflammation [
33,
36,
50]. Proof-of-concept studies using a cathepsin-activatable probe for
in vivo imaging of protease activity associated with RA animal models have also recently been published [
36]. A number of issues remain to be clarified, however, including the utility of optical tomography to provide quantitative readouts that discriminate between DMARD and non-DMARD treatments, and a clear and comprehensive comparison of imaging readouts with standard measures of clinical score and paw swelling.
For the first time, we show the benefit of optical tomographic imaging in a mouse model of arthritis and demonstrate that tomography not only can provide a quantitative measurement of disease severity but also can accurately define the scope of disease in the ankle joints versus interphalangeal joints of the hindpaws (Figure
5). The higher signal in the ankle joints probably indicates a greater magnitude of disease rather than a significantly greater severity of disease processes, as more total disease activity would obviously be occurring in the larger ankle joint. The multiplex approach to quantifying two or more biomarkers correlated with the histopathology, and the fluorescence data could be used to accurately predict histology inflammation scores prior to collection of tissue. These results confirmed that imaging of cathepsin activity, MMP activity, and bone turnover is a successful non-invasive diagnostic modality, capable of providing robust measures of disease progression.
Optical tomographic imaging of NIR imaging agents in CAIA has the potential to be used in drug discovery research by virtue of deep tissue penetration, quantitative readout (picomoles rather than light intensity), and the pairing with NIR imaging agents that detect the cellular participants in the underlying disease pathology. Both p38 MAPK inhibitors and COX-2 inhibitors have previously been shown to effectively reduce clinical signs of disease and paw swelling measurements in rodent models of arthritis [
6,
7,
51,
52]. COX-2 inhibitors and other NSAIDs are better at providing symptom relief than at altering disease progression [
10,
11], however, whereas p38 MAPK inhibitors significantly decrease underlying inflammation and bone destruction [
4‐
8]. We confirmed these observations, showing an overestimation of efficacy of celecoxib and p38 MAPK inhibitors on clinical score and paw swelling as compared with effects by histological assessment (Figures
1 to
3).
A large panel of plasma biomarkers (including IL-6, G-CSF, eotaxin, and MMP-3) also failed to accurately characterize therapeutic efficacy, with both overestimation and underestimation of observed responses (Figure
4) depending on the specific biomarker and therapeutic agent. Although it is likely that the failure of many of the tested biomarkers was due to the very acute nature of this model, it does highlight the disconnection, or perhaps delay, of plasma biomarkers relative to vigorous onset of joint inflammation and destruction. This very failure of plasma biomarkers further suggests that direct imaging of the site of disease should generally be a more accurate means of assessing chronic progression as well as acute flares of disease in human RA. In contrast, imaging with either ProSense750, MMPSense680, or OsteoSense680 provided an excellent discrimination between p38 MAPK inhibitor and celecoxib therapies (Figures
6,
7 and
9), and correlated extremely well with histology inflammation scores (Figure
3a) and human therapeutic observations. In preliminary studies, we found a similar dissociation between standard measures compared with imaging and histology in mouse collagen-induced arthritis when mice were imaged on day 50 post immunization (JD Peterson and TP Misko, unpublished observations).
Imaging CAIA mice using OsteoSense680 appears to be a rapid and sensitive means of detecting changes in bone turnover associated with disease progression. It is important to note that bone resorption was mild in this acute model, as assessed by histopathology (Figures
2 and
3), yet OsteoSense680 showed a twofold increase in signal on day 8 and a fourfold increase by day 15, suggesting significant bone changes were occurring. Furthermore, this readout clearly differentiated between p38 MAPK inhibitor and prednisolone as compared with celecoxib on day 15. It is interesting to note that prednisolone decreased the OsteoSense680 signal on day 8 to the level of the control mice; however, on day 15 the signal increased above the control in the absence of any apparent disease. The longer treatment time with prednisolone probably induced some bone loss, as we have seen previously that this dose of prednisolone will indeed cause some bone loss (and increased OsteoSense680 incorporation) in normal mice (JD Peterson and R Rader, unpublished observations).
Competing interests
KOV, SK, and JDP are employees of VisEn Medical. TPL, MM, RR, JTL, MAA, and TPM are employees of Pfizer Global Research & Development (St Louis, MO, USA). Funding of these studies was shared in a collaboration between Pfizer Global Research & Development and VisEn Medical. The research documents the utility of VisEn imaging agents and imaging technology in addressing specific biological questions in arthritis, but VisEn receives no direct financial gain as a result of publication. Pfizer employees have no financial stake in VisEn Medical.
Authors' contributions
JDP designed, analyzed, and provided oversight for all in vivo imaging studies, KOV performed the in vivo studies. SK carried out preliminary in vivo validation studies. TPL assessed disease pathology in histologic sections and participated in drafting the manuscript. RR made substantial contributions to study conception. TPM helped in the design and interpretation of the studies and participated in drafting the manuscript. JTL performed the cytokine and chemokine multiplex analysis. MAA performed all MMP-3 assays. MM provided valuable input on arthritis study design and prepared drug formulations.