A window on disease pathogenesis and potential therapeutic strategies: molecular imaging for arthritis

Novel molecular imaging techniques are at the forefront of both preclinical and clinical imaging strategies. They have significant potential to offer visualisation and quantification of molecular and cellular changes in health and disease. This will help to shed light on pathobiology and underlying disease processes and provide further information about the mechanisms of action of novel therapeutic strategies. This review explores currently available molecular imaging techniques that are available for preclinical studies with a focus on optical imaging techniques and discusses how current and future advances will enable translation into the clinic for patients with arthritis.

Th e objectives of molecular imaging are the visualization, characterization and quantifi cation of molecular and cellular processes non-invasively within intact living organisms. Th is can help evaluate physiological and pathophysiological processes, facilitate diagnosis and monitor the eff ects of therapy. At the preclinical stage novel molecular imaging techniques can facilitate the development of new therapies and understanding of novel mechanisms of action of biologically targeted agents.
Rheumatoid arthritis (RA) is a chronic systemic infl ammatory disease, primarily characterised by infl amed synovial tissue in multiple joints leading to localised destruction of bone. Despite signifi cant advances in conventional imaging strategies, such as the use of power doppler ultrasound scans and magnetic resonance imaging (MRI), the early diagnosis and monitoring of infl ammatory conditions such as RA remains challenging. Current imaging refl ects irreversible pathological and anatomical change as opposed to perturbations in specifi c molecular pathways. Pathological change visualized on X-ray imaging may not be seen until many months after disease onset [1]. Osteoarthritis (OA) is characterised by joint pain, inactivity-related stiff ness, impaired social role and reduced quality of life, which may be associated with radiographic abnormalities. It is the most prevalent joint disease and a major cause of disability [2][3][4]. While in some ways similar to RA, it is recognised that cartilage loss may occur for several years before even minor changes can be detected on plain radiographs in OA [5]. Molecular in vivo imaging in animal models of disease is also important in increasing our understanding of disease pathogenesis and in developing methods of monitoring disease activity in vivo. Robust disease monitoring will also allow for better appraisal of potential therapeutics. Ultimately, the aim is to translate molecular imaging techniques into functional systems for imaging of human diseases such as RA and OA.

Imaging modalities and their application to musculoskeletal disease
Radionucleotide imaging: the past and the future Th e following sections detail how diff erent radio nucleotide imaging modalities have been utilized in musculoskeletal imaging practice and how they may be further deployed.
Radionucleotide techniques have long been established for imaging in preclinical animal models as well as in RA and related autoimmune conditions. A variety of methods have been employed successfully for imaging leukocytes with 99m Tc, 67 Ga and 111 In, with particular focus on their use in RA, where they remain in routine clinical use as diagnostic imaging techniques [6]. Th e oldest radiopharmaceutical proposed for imaging infl ammation was 67 Gallium citrate and this remains an established technique for imaging pulmonary and musculoskeletal infl ammation, especially in sarcoidosis [7]. A radionucleotide can also be applied to a macromolecule making use of the permeability change that is recognised to occur around sites of infl ammation [8]. Th ere are numerous examples of so-called non-targeted approaches, including radiolabelled liposomes, dextran, nanocolloid and human immunoglobulin [9][10][11][12]. While all these techniques have a low cost and are widely available, resolution and sensitivity are relatively low. As a further example, 99m Tc-oxidronate (Tc-HDP) displays abnormal uptake over both currently infl amed and chronically damaged joints and is very sensitive for the detection of joint and subchondral bone abnormalities [13,14], but it cannot distinguish accurately between actively and chronically infl amed joints [15].
A radionucleotide can be applied to a specifi c molecular target to improve these issues. mAbs with high specifi city and high affi nity for their target antigens can be utilized for delivery of agents, including radio nucleotides, enzymes, drugs, or toxins, in vivo. Th e Fab is a region on an antibody that binds to antigens. It is composed of one constant and one variable domain of each of the heavy and the light chain. Th is may benefi t both targeting specifi city, due to decreasing the size of labelled conjugate, and may also reduce immunogenicity as described below. Scintigraphy utilizing a 99m Tc-anti-Eselectin-Fab has been used with good eff ect to image synovitis in patients with RA, in this case demonstrating improved specifi city compared to a conventional tracer for bone and joint (Tc-HDP). Th is technique demonstrated particular specifi city for targeting active joint infl ammation [16]. Th e absence of signal uptake in normal joints by E-selectin targeted imaging described above allows for markedly improved detection of active joint disease in RA. Th e mAb described in this study is of murine origin, which raises the possibility of host immunogenicity. However, the small amounts of mAb required for imaging and also the use of the Fab fragment devoid of Fc portions, which are thought to be responsible for generating host immunity, may reduce the likelihood of a human anti-mouse antibody (HAMA) response. In this study patients were not tested for a HAMA response, but some were imaged again with 99m Tc-anti-E-selectin-Fab 2 to 4 months later with no detectable change in the biodistribution of labelled antibody. In a previous study of 14 patients with RA that received 111 In-labelled F(ab') 2 fragment of E-selectin, HAMA responses were measured by ELISA comparing blood samples taken pre-immunisation and postimmuni sation (14 days). No HAMAs were detectable following injection.
Other specifi c molecular targeting techniques have been developed but can be aff ected by the non-specifi c localization of agents to infl amed tissues. An example includes 111 I-labelled IL-1 receptor antagonist (IL-1ra), which did not show increased localisation compared to radiolabelled albumin [17]. It would appear in this study that at the doses of labelled IL-1ra used, the most substantial eff ect was for local changes to refl ect nonspecifi c traffi cking of the molecule due to local permeability changes in areas of infl ammation. Th e therapeutic effi cacy of IL-1ra (anakinra, which is licensed for treatment in RA) requires repeated administration at 1,000-fold the doses used in the above imaging study. If imaging was undertaken with much higher concen trations of labelled anakinra, greater specifi city of signal compared to radiolabelled albumin may be achieved; however, increasing the dose of imaging agent to that level would lead to unacceptable levels of radiation.
Th e scintographic detection of radiolabelled TNF mAb has also been investigated. Twenty-fi ve percent of the 99m Tc labelled anti-TNF signal could be abrogated following prior administration of unlabelled anti-TNF [18]. Th is demonstrates a specifi c component to imaging, but also suggests that 75% of the signal was non-specifi c accumulation of labelled tracer at sites of infl ammation.
Interestingly, in RA, other workers have demonstrated that 99m Tc-anti-CD3 scintigraphy can be used to diff er entiate between autoimmune rheumatic diseases and strongly correlates with clinical fi ndings for patients with RA [19]. More recently it has been demonstrated that anti-CD3 therapy is highly eff ective for the treatment of collagen-induced arthritis (CIA), a murine arthritis model, with the capacity to generate anti-arthritic CD8+ regulatory T cells (Tregs) and expand the relative numbers of CD4+ Tregs [20]. Th is highlights the potential of how both imaging and therapy could be synergised.
To determine whether overexpression of the high affi nity folate receptor on activated macrophages can be utilised to selectively target imaging agents to sites of infl ammation in rats with adjuvant-induced arthritis, folic acid was conjugated to a 99mTc chelator and its distribution was visualised using gamma scintigraphy in healthy rats, rats with adjuvant-induced arthritis and arthritic rats that had been depleted of macrophages [21]. Uptake was also confi rmed by pre-administering unlabelled folate -this demonstrated that the folate imaging conjugate concentrated in the extremities of diseased rats but not in the extremities of healthy rats. Depletion of macrophages also reduced tissue folate receptor content with a concomitant reduction in the folate targeted signal. Th is study demonstrates that cell based labelling may also be viable for both functional imaging and targeted delivery of therapeutics [21].

Positron emission tomography and single photon emission tomography
Positron emission tomography (PET) imaging is based on imaging emission from isotopes such as 14 F and 64 Cu.
Th ese decay and emit positrons that on collision with electrons emit λ-rays at 180 degrees to each other. Th ese are detected when they reach a scintillator in the scanning device, creating a burst of light that is detected by photomultiplier tubes or silicon avalanche photodiodes.
PET imaging has been applied in in vivo models of arthritis. For example, in the K/BxN mouse model of RA, the transfer of autoantibodies specifi c for glucose-6phosphate isomerase (GPI) into naïve mice rapidly induces joint infl ammation similar to that seen in RA. Th e autoimmune target for the KRN T-cell and antibody response is GPI, which is involved in glycolysis and is ubiquitously present within the cytosol and within tissue at low levels. More than 50% of humans aff ected with RA have elevated levels of anti-GPI antibodies. In the K/BxN mouse model, Wipke and colleagues [22] have shown by PET imaging that purifi ed anti-GPI IgG localizes specifi cally to the front and rear distal joints within minutes of intravenous injection, reaches saturation by 20 minutes and remains localized for at least 24 hours. By contrast, control IgG does not localize to joints or cause infl ammation. For this study anti-GPI antibodies were labelled with 64 Cu and injected into the recipient mice, which were then imaged by micro-PET [22]. Th is work has been extended by Binstadt and colleagues [23], who have gone on to demonstrate by intravital imaging that anti-GPI antibodies caused macromolecular vaso permea bility localized to sites destined to develop arthritis. Th ese studies highlight how molecular in vivo imaging techniques can be used to dissect the dynamic and organ specifi city of autoimmune attack.
Human studies have also been undertaken. 18 F fl uorodeoxyglucose ( 18 F-FDG) PET imaging and methyl- 11 C-choline, an isotope of choline that is incorporated into cell membranes, can be used to image cellular proliferation. Roivainen and colleagues [24] compared 18 F-FDG PET and methyl-11 C-choline PET with MRI measures of synovial proliferation at the knee in a study of ten patients with infl ammatory arthritis including RA. Th ey found that synovial volumes measured on contrastenhanced T1-weighted MRI scans were highly correlated with PET measures of membrane turnover and glucose metabolism, including the standardized uptake value (SUV) for these isotopes (r values of 0.834 for 11 C-choline and 0.714 for 18 F FDG).
Single photon emission computed tomography (SPECT) is a nuclear medicine tomographic imaging technique that utilizes the emission of gamma rays. SPECT imaging is performed by using a gamma camera to acquire two-dimensional images from multiple angles. Th is is reconstructed into a three-dimensional dataset, similar to those obtained from other tomographic techniques such as MRI, computed tomography (CT), and PET. SPECT is similar to PET in its use of radioactive tracer material and detection of gamma rays. In contrast with PET, however, the tracer used in SPECT emits gamma radiation that is measured directly. SPECT scans are signifi cantly less expensive than PET scans, in part because they are able to use longer-lived, more easily obtained radioisotopes than PET. Th e widespread use of PET is limited by a shortage of centres able to generate positron radioisotopes that also have short half lives. SPECT uses isotopes that are already in widespread use, such as 111 In and 99 mTc. Rapid advances in SPECT, including the use of multiple pinhole collimators, allow high-resolution imaging to less than 1 mm. Th is enables mouse physiology to be quantifi ed. Th is has been used to good eff ect in a humanized mouse model of arthritis. NanoSPECT/CT co-registration technology has been used to image the uptake of 111 In labelled anti-E-selectin antibody into human synovium that was transplanted into SCID mice following graft stimulation with TNFα. Th is demonstrated that activity could be detected in the graft at multiple time points following injection with clear delineation of the trans plants in reconstructed images with signifi cant diff er ences to 111 In-labelled isotype control antibody [25]. Th ese results have demonstrated that this may be a technique with signifi cant promise for small animal imaging. A translational study of high resolution SPECT imaging in a murine arthritis model and in patients with RA co-registered with MRI has also been performed and this is described below.

MRI and high-resolution computed tomography
MRI scanning is the current gold standard modality in clinical practice for imaging synovitis and tenosynovitis in patients with infl ammatory arthritis. Studies have shown correlations between the degree of infl ammation and the vascularity of synovium obtained at biopsy and post-contrast enhancement on matching dynamic MRI scans. Scoring systems have been devised that are based on quantifying synovial membrane thickening and signal intensity on static post-contrast scans. Moderate to high reliability has been demonstrated with trained readers and quantifi cation of synovitis in this way is being used increasingly as an outcome measure in clinical trials to assess response to therapy. MRI-observed synovitis is almost invariable in those with active RA, but recent studies have also demonstrated its presence in patients in clinical remission, emphasizing the sensitivity of this technique and the importance of subclinical joint infl ammation [26]. Synovial infl ammation determined by MRI has been shown to be common in OA [27]. However, MRI studies are complex and have a new range of variables, including diff erent imaging sequences, protocols and hardware, but quantifi cation is improving, allowing for a good level of accuracy and reliability in measurements of cartilage volume and thickness [28].
For example, a novel method of measurement of focal cartilage thickness change by MRI by detailed mapping of changes in cartilage thickness has been developed. Th is demonstrated marked focal changes in cartilage thickness over time and correlated with disability [29].
High-resolution delayed gadolinium-enhanced MRI of cartilage (dGEMRIC) is an alternative technique that can provide further information on the composition and structure of the cartilage matrix. Cartilage consists of relatively few cells and a hydrated extracellular matrix consisting mainly of proteoglycans and collagen, which are responsible for load distribution and compressive stiff ness. Proteoglycans consist of numerous glyco samino glycan (GAG) side chains that are negatively charged, providing the extracellular matrix with the majority of its fi xed charge -this is referred to as the fi xed charge density (FCD). An early feature of OA is the loss of GAGs and hence FCD from the tissue; T1 mapping of cartilage is performed after intravenous administration of gadolinium-based contrast agent, which is able to provide an indirect measurement of GAGs, and hence cartilage degradation, by detecting changes in the FCD [30][31][32]. In patients with knee pain, GAG depletion as measured by a dGEMRIC index has been shown to be predictive of development of knee OA in a 6-year follow-up. Th is technique may therefore act as a clinically relevant measure of cartilage integrity as well as provide information about underlying molecular perturbation [33].
Cartilage quality assessment in small joints may also be made using dGEMRIC. An initial study examined four patients with OA of the fi rst carpometacarpal joint, demonstrating diff erences between asymptomatic and OA joints [34]. dGEMRIC may also help to assess cartilage degeneration in the metarcarpal phalangeal joints of patients with RA. A study of 31 metarcarpal joints in 10 patients with RA demonstrated signifi cant changes in signal from the second and third metarcarpal joints when compared to control subjects, whereas measurements of cartilage thickness were comparable in both groups [35].
Custom built high-resolution MRI devices have been developed to image small joint structures, including the bone trabeculae, cartilage, ligaments and tendons [36,37]. Th ese have been developed to generate detailed anatomical assessment of the earliest stages of OA [38]. Th is has also added to understanding of the relative anatomical diff erences between OA and changes in spondylo arthritides such as psoriatic arthritis. Advances in imaging technology, such as high-resolution 'microscopy' MRI and whole body MRI, and improved protocols, such as ultrashort echo time, are expected to further delineate pathophysiology [39].
Paramagnetic contrast agents have also been utilised for cell tracking studies to monitor T-cell homing in vivo after loading T cells with superparamagnetic iron oxide (CLIO) nanoparticles conjugated with a peptide sequence from the transactivator protein (Tat) of HIV-1. T cells were isolated from C57BL/6 (B6) mice and homing of T cells to the spleen was demonstrated by MRI following their re-injection. A similar technique was also used to image T-cell infi ltrates in the pancreas of diabetic mice [40]. Both studies highlight the potential of MRI to track cellular events in deep tissues within living animals, although this technique may be hampered by the minimum number of cells that can be visualised during imaging. Ultrasmall superparamagnetic iron oxide (USPIO) nanoparticles are an MRI contrast agent that consists of an iron oxide core of about 5 nm in diameter that is surrounded by dextran, which increases the diameter to about 30 to 50 nm. Several USPIO preparations are in advanced stages of clinical trials, and their safety in humans has been increasingly established [41,42]. Conjugation of USPIO nanoparticles with ligands off ers the possibility of MR imaging of molecular targets. Reynolds and colleagues [43] have demonstrated that in vivo expression of E-selectin during endothelial activation in an infl amed mouse ear model can be depicted with MR imaging and that the conjugate did not accumulate in non-infl amed endothelium.
Microfocal CT (micro-CT) is a three-dimensional high-resolution CT technique that allows volumetric assess ment at peripheral skeletal sites. It can provide accurate depiction of structural changes in arthritides. For example, periarticular bone structure in RA patients was compared to healthy individuals using this technique and demonstrated exact detection of morphologic changes of juxtaarticular bone in healthy individuals and RA patients [44]. While not directly a functional imaging technique, this may hold promise as an anatomical coregistration modality with other forms of imaging so that anatomical and molecular change can be measured during the time course of arthritis.

Optical imaging: a viable alternative molecular technique
Optical imaging has already been developed for in vitro and ex vivo applications in molecular and cellular biology (for example, fl uorescence confocal microscopy), and there are now signifi cant advances in its development as a whole animal in vivo imaging technique, particularly in the context of arthritis [45]. Th e future holds considerable promise for non-invasive visualisation of specifi c molecular targets, which as with the other imaging techniques presented here can be synergised with other imaging modalities such as CT or MRI. Th e development of near infrared (NIR) fl uorophores has also off ered advan tages for in vivo imaging. Signifi cant tissue autofl uorescence occurs at lower wavelengths and decreases signal sensitivity, whereas this is reduced in the NIR range. Hence, NIR in vivo imaging may off er a unique window onto disease pathogenesis [46]. By contrast to the other techniques presented, it is comparatively low cost, does not require radiation, and multiple images can be acquired rapidly. Th e following sections detail how optical imaging may be utilised to visualise arthritic conditions in vivo.

Non-specifi c imaging agents detect permeability change in infl ammatory states
Infl ammatory arthritides such as RA are characterised by vascular infl ammation and pathologic neoangiogenesis [47,48]. Disordered new vessels have increased vascular leakiness to macromolecules compared to normal vessels [8]. As the vascular permeability increases, this leads to plasma extravasation, oedema and swelling. Th is local permeability change has been utilised to explore the nonspecifi c uptake and deposition of fl uorescent dye within the arthritic joint in vivo. Enhanced localised deposition of free dye (not bound to a disease-specifi c antibody or other target) in the arthritic joints of animals with induced arthritis have been examined in several studies. Th e NIR fl uorophore Cy5.5 was visualised in arthritic knee joints following induction of antigen-induced arthritis in mice [49], and demonstrated an approximately 1.5-fold increase in signal in arthritic compared to nonarthritic joints. Two other NIR dyes were evaluated in a murine Borrelia-induced Lyme arthritis model [50]. Th ese techniques rely on the generalised increased blood perfusion and 'vascular leakiness' that are recognised to be present around the infl amed joint [51]. Th is may lead to as much as a 40-fold increase in macroglobulin permea bility in the infl amed joint [8]. A number of diff erent approaches have been deployed to make use of vascular permeability change in order to improve the therapeutic index of drug treatment. One possibility is the direct conjugation of a drug to another larger molecule -an example of this is conjugation to polyethylene glycol (PEG). Th is approach has been successfully deployed for a number of drugs in routine clinical use, such as the PEGylated forms of interferon-α and doxorubicin for the treatment of hepatitis C and breast cancer, respectively [52]. In RA this approach has demonstrated improved localisation to sites of infl ammation by measuring increased levels of fl uorescence in infl amed paws with a fl uorescently labelled PEGylated anti-TNFα agent compared to a non-PEGylated TNFα inhibitor [53]. Th e pegylated anti-TNFα agent Certolizumab is licensed for clinical use in patients with active RA [54].

Targeted imaging agents Antibody-ligand interactions: determining specifi city of molecular targeting
Specifi city of dye traffi cking may be enhanced by utilising an antibody, or its fragment, directed at a specifi c molecular target. Specifi c deposition of dye in the arthritic joint of antigen-induced arthritic mice has been achieved by targeting macrophages with Cy5.5-labelled anti-F4/80 antigen. Th e F4/80 antigen is expressed on the macrophages when they accumulate in infl amed joints [55]. Th ere was, however, a small increase in signal in the uninfl amed contralateral knee joints as well as increased uptake in infl amed knee joints with the Cy5.5-labelled isotype control antibody. Th is highlights the challenge of identifying specifi c versus non-specifi c deposition of dye within the infl amed joint. Optical imaging of arthritis utilising a probe directed against folate receptors on activated macrophages demonstrated a 2.3-fold increase in signal in infl amed compared to uninfl amed joints [56]. Th is increase was greater than that obtained following injection of free dye. Another study has examined in vivo fl uorescent signal returned from allogeneic leukocytes that were labelled ex vivo with fl uorescent dye and reinjected into arthritic mice [57]. Steroid reduced the signal intensity obtained; however, the degree of quantitative analysis was hampered by a relatively small change in signal intensity compared to the background signal.
In our laboratory, we have demonstrated that E-selectin-targeted in vivo imaging is a quantifi able method of detecting endothelial activation in arthritis and can potentially be applied to quantify disease and investigate the eff ects of novel therapies [58]. Figure 1 demonstrates that E-selectin-specifi c signal can be detected in collagen-induced arthritis, a mouse model of RA. Signifi cant diff erences in anti-E-selectin antibody NIR fl uorescent signal can be detected compared to signal received from an isotype control antibody. In addition, it is also possible to co-register the fl uorescent signal with digital X-rays of the aff ected area taken concurrently. Th is technique has potential to be translated into a method of delineating the changes occurring to activated endothelium in health and in disease both prior to and following therapy.

Activatable imaging agents enable signal amplifi cation
Amplifi cation of fl uorescent signal at the site of infl ammation may be possible by using protease activatable probes. Here a cleavable peptide sequence is linked to fl uorescent molecules that are quenched when in close proximity to each other, but unquenched following cleavage by local proteolytic activity [59]. Th is may be particularly useful for optical imaging of OA, as local perturbations in proteolytic activity may be comparatively small. Cathepsins have been associated with arthritis and bone degeneration [60]. A cathepsin B activatable NIR fl uorescent probe was used in a mouse model of OA and showed a three-fold diff erence in signal intensity between normal and osteoarthritic joints [61,62]. Th is particular probe has also been applied to tumour detection [63] as well as the localisation of atherosclerotic plaques in vivo [64]. A similar probe was used to demonstrate that methotrexate could abrogate some of the increased signal found in arthritic joints [65]. A further study also demonstrated that specifi c signal could be reduced and correlated with a decrease in clinical disease indices such as redness and measurable paw swelling [66]. However, these studies did not examine whether an uncleavable form of peptide demonstrated similar signal or whether any specifi c inhibitor of the protease could prevent cleavage in the arthritic joint.
At the preclinical stage a number of diff erent advances in generating either fl uorescence (for example, GFP expression) or light (bioluminescence) have become powerful tools to delineate disease pathogenesis over multiple time points. Bioluminescence is the conversion of chemical energy to light in living organisms, most commonly under control of the luciferase enzymatic system. Th ese enzymes convert the luciferin substrate in the presence of several co-factors and adenosine triphosphate to generate oxyluciferin and light [67]. Th is technology has been most widely used in reporter gene assays, where the promoters of genes under study have been linked to the luciferase gene, and the resultant light is a direct measure of promoter activity. Because the luciferase enzyme needs to be transfected into cells or expressed in a transgenic line, its use is limited to animal models. Pioneering studies have demonstrated in vivo imaging of bone forming osteoblasts. Gazit and colleagues [68] generated a transgenic mouse model containing the luciferase promoter under control of an osteoblast-specifi c osteocalcin promoter fragment. In this study, light emission correlated well with the activity of osteoblasts in parts of the bony skeleton, matching histological expression of osteocalcin expression and quanti fi cation by real time PCR. Furthermore, in two skeletal repair models, fi rstly by bone fracture and secondly by marrow ablation, quantitative measurements could be made in vivo [68]. Bioluminescence has also been used successfully to image and study animal arthritis models. Transgenic mice that express luciferase under control of the nuclear transcription factor NF-κB have enabled real time imaging of NF-κB expression during the course of an arthritis model in intact animals [69]. Imaging of arthritic joints revealed a strong activation of NF-κB on day 6 post-arthritis induction and increased expression over the following 4 days. One particular advantage bioluminescence has over traditional fl uorophores is that tracking can be monitored over longer periods (days to weeks) [70].

Image co-registration techniques combine anatomical and molecular data
Th e potential application of molecular imaging techniques is determined by the ability to both quantify sensitivity and specifi city and map this signal to discrete tissue structures in vivo. Pre-clinical and clinical multiple imaging platforms have been developed, such as PET-CT, SPECT-CT or PET-MRI. For example, PET-MRI may allow for the analysis of one or more functional processes simultaneously along with the delineation of highresolution morphology [71]. Hybrid cameras that combine SPECT and helical CT can be used to correlate scintographic information with morphologic information The corresponding quantifi cation of MFI at 8 hours for either healthy (non-arthritic) animals or mice with clinically defi nite arthritis and paw thickness ≥2.2 mm. Background fl uorescence levels from uninfl amed paws have been subtracted. Data are mean ± SEM, and were analysed by one-way ANOVA versus arthritic anti-E-selectin-injected mice: **P < 0.01, ***P < 0.001. Reproduced with permission from [58].
in one session. Use of SPECT/CT hybrid imaging may signifi cantly increase the diagnostic accuracy of skeletal scintigraphy in the staging of malignant disease but it has been little used for the interpretation of musculoskeletal disorders. A recent pilot study has demonstrated the potential benefi ts of this technique for improving diagnostic accuracy in diff erentiating the aetiology of musculoskeletal disorders -among 34 lesions classifi ed as OA on planar and SPECT imaging, 7 were reclassi fi ed as fracture and one as a benign tumour [72]. Th is encourages further follow-up work comparing fi ndings with other imaging modalities such as MRI, clinical outcomes and, where possible, histo pathological correlates.
A combinatorial approach has also been developed for optical imaging co-registered to MRI. Concurrent MRI and diff use optical tomography of the breast after indocyanine green enhancement has been undertaken [73]. Th is type of technology has developed in conjunction with suitably constructed probes. For example, Olson and colleagues [74] have developed activateable cell penetrating peptides linked to nanoparticles to serve as dual probes for in vivo fl uorescence and MRI of proteases. Th is study reports on the use of in vivo visualization of matrix metalloproteinase activities by MRI and fl uorescence of dendrimeric nanoparticles coated with activateable cell-penetrating peptides that have the potential to be labelled with both Cy5 and gadolinium. Uptake of such nanoparticles in tumours is 4-to 15-fold higher than for unconjugated activateable cell-penetrating peptides in a mouse tumour model. With fl uorescent molecules, residual tumour and metastases as small as 200 μm were detected. In addition, once activated, the Gd-labelled nanoparticles deposit high levels (30 to 50 μM) of Gd in tumour parenchyma, with even higher amounts deposited in regions of infi ltrative tumour, resulting in useful T1 signal contrast lasting several days after injection.
Infl ammatory arthritic lesions have been imaged in experimental arthritis and in patients with RA using a high-resolution multi-pinhole (MPH)-SPECT technique [75]. In the human studies MPH-SPECT images depicted a detailed visualization of tracer accumulation in the bony structures of the hand and fi nger joints; these studies were also capable of imaging increased bone metabolism that had appeared normal with other imaging modalities. Th is is demonstrated in Figure 2. Ostendorf and colleagues [76] have also demonstrated that MPH-SPECT is sensitive to early changes in RA and OA and that the diff erences between them can be detected by diff erent patterns of uptake. It is also interesting to note that this study used a conventional tracer, so it is possible that much greater sensitivity could be achieved by conjugating the tracer to molecules directed at specifi c molecular targets.

Future perspectives
Conventional and currently clinically available imaging strategies for detecting and monitoring arthritis have relied mainly on detecting the endpoints of autoimmune pathological processes. But molecular imaging now off ers a new way of detecting and monitoring disease. Developments in radionucleotide imaging discussed here will result in methods of detecting radiolabelled agents directed at specifi c molecular targets. Th ese have the advantage that many of the facilities are already in clinical practice. Further developments in software will mean that novel co-registration techniques will enable precise anatomical localisation of molecular change, but exposure to radiation, especially if multiple imaging points are required, may limit routine clinical use. Similarly, MRI scanning is in widespread clinical use and the development of novel imaging protocols and validation and qualifi cation of their use means that these techniques can readily be translated into clinical practice. However, there is often heavy clinical demand for current scanning equipment and imaging can be time consuming. Furthermore, there may be issues related to the toxicity of contrast agents; it is well recognised that gadolinium contrast agents may be nephrotoxic [77], and superoxide particles may have eff ects on endothelial cells [78,79]. Optical imaging off ers a viable alternative imaging strategy; images can be obtained rapidly and devices are relatively inexpensive. Furthermore, there is no exposure to radiation. Developments in fl uorophore technology have enabled accurate and quantifi able measurement of molecular processes. Th is technique is also readily combined with other imaging strategies.
While it is recognized that joint damage is related to disease activity in RA, the degree of progression and the temporal associations between the two are unclear [80]. However, progression of joint destruction is not always directly coupled to the signs and symptoms of RA. Following anti-TNFα therapy it has been noted that joint disease can be retarded even if active disease is ongoing [81,82]. Conversely, there are also reports that joint damage can accrue even in states of remission [83,84]. Using molecular techniques to sensibly determine sub clinical levels of joint infl ammation due to the upregu lation of specifi c molecular pathways may be one way of determining levels of disease activity and response to therapy. Th is was explored in the study by Brown and colleagues [84] where the long-term signifi cance of sub clinical synovitis was assessed in relation to overall structural outcomes. Th is was performed by analyzing musculo skeletal ultrasonography in addition to MRI scoring measures. Targeted functional imaging therefore off ers a method of assessing disease parameters in a diff erent way.
Rather than focusing on pathological outcome, changes in specifi c molecular pathways can be measured. Many of the studies presented here have utilised an antibody targeted approach. Th ere are several shortcomings of mAbs, including their immunogenicity, suboptimal target ing and pharmacokinetic properties, and practical issues of production and labelling. Genetic engineering provides a powerful approach for redesigning antibodies for use in oncologic applications in vivo. Recombinant fragments have been produced that retain high affi nity for target antigens, and display a combination of high targeting specifi city with concomitant clearance from normal tissues and the circulation in animal models. Synovium-specifi c peptides can also be generated by in vivo phage display selection. Th ese may uniquely hone to sites of infl ammation within the infl amed synovium [85]. Clustering of fl uorophores onto a larger scaff old may also be used to amplify the fl uorescent signal.
Aptamer-based technologies and nanoparticles are potential novel particles that could be deployed successfully for molecular imaging. As oligonucleotide ligands, aptamers are comparable to antibodies in specifi city and affi nity for their target molecule, typically a protein [86,87]. At 8 to 15 kDa, aptamers are intermediate in size between antibodies (150 kDa) and small peptides (15 kDa) and are slightly smaller than sFv (25 kDa). As polyanions, aptamers are quite diff erent in composition from sFv. As synthetic molecules, aptamers readily support site-specifi c modifi cations that maintain structure and activity. Aptamers can be coupled to diagnostic or therapeutic agents and to bioconjugates, such as PEG polymers, that can alter aptamer pharmacokinetics. Th e parameters for the in vivo activity of aptamers have been investigated in rats with an aptamer directed against L-selectin [88]. Previous therapeutic work with aptamers has focused on blocking protein function; by far the most advanced work is represented by the use of a vascular endothelial growth factor aptamer, pegaptanib sodium (Macugen; Pfi zer and Eyetech) [89], now approved for treatment of macular degeneration [90]. Th e small size and polyanionic nature of aptamers may lead to rapid blood clearance and tissue uptake and may minimize the residence in liver and kidney, providing some potentially useful features for imaging and radiotherapy. Initial experiments to address aptamer suitability for in vivo imaging have been reported [91,92]. Aptamers (small oligonucleotide ligands) combined with small particles called nanorods made out of gold and silver act as a platform to accommodate multiple weak binding reactions. Fluorescently labelled aptamers increase their affi nity for diff erent cancer cell lines by 26-fold, resulting in a 300-fold increase in fl uorescence signal compared to conventional techniques [93]. Th e use of nanorods in this instance may also open potential for therapeutic options. Since the nanorods have strong infrared absorption, hyperthermia may be generated locally, which could specifi cally ablate tumour or neoangiogenic infl ammatory tissue.
Th ere are other examples of merging both imaging and intervention. In the adjuvant arthritis model of RA, folate labelled with fl uorescein and conjugated with haptens (small immunogenic molecules) was used to identify activated macrophages and deliver immunotherapy to the activated cell population. In these experiments, fl uorescein isothiocyanate (FITC)-labelled folate was administered orally to folate-depleted animals and taken up preferentially by these cells as a result of their high surface expression of a folic acid receptor [21,94]. Th e haptens attached were subsequently expressed on the macrophage cell surface, making them highly immunogenic. Induction of anti-hapten antibodies in these animals via a vaccination schedule resulted in selective depletion of activated macrophages, with an anti-infl amma tory eff ect comparable to that achieved by methotrexate, etanercept and anakinra in this model. Th is kind of approach could be used to target a wide variety of cells with selectively toxic agents without exposing the patient to the risks of non-specifi c immunosuppression.
GFP from the hydroid jellyfi sh Aequorea victoria and GFP mutants and homologues are used intensely in biotech nology. Th ey represent the only available fl uorescent tag fully encoded in a single gene. GFP-like proteins are utilized for in vivo labelling of organisms, cells, organelles and proteins, as well as in molecular sensors that probe diff erent aspects of the intracellular environment. Th e excita tion and emission spectra of these fl uorescent reporter proteins used thus far for bioluminescence imaging has not exceeded 598 nm and 655 nm [95]. Th is signifi cantly hampers the penetration of excited light through tissue and therefore aff ects the resolution of images that may be obtained. More recently, however, mammalian expression of infrared fl uorescent proteins has been engineered from a bacterial phytochrome. Th ese higher wavelengths penetrate tissue well and may be extremely useful for further in vivo imaging studies [96]. Combined with tomographic imaging techniques, localisation may be made to precise anatomic sites. Most recently a novel multispectral opto-acoustic imaging technique has shown signifi cant potential to image in vivo to resolutions of 20 to 100 μm with depth penetration that could range up to several centimetres. Combining the signifi cant enhancements in reporter proteins suitable for in vivo use and novel hardware solutions holds signifi cant promise [97].

Conclusion
Th is review has demonstrated that functional molecular imaging techniques have the potential to visualise pathobiology in a new way. Th ey can act to delineate the mechanism of action of therapy and can also be used to monitor therapeutic outcomes in patients with arthritis and other infl ammatory autoimmune conditions, holding much promise as future imaging modalities.

Competing interests
The authors declare that they have no fi nancial or non-fi nancial competing interests.