Molecular imaging of inflammation-Current and emerging technologies for diagnosis and treatment

Available online 20 April 2020


Definition of molecular imaging
Molecular imaging concerns the visualization, detection and quantification of physiological and pathologic/disease processes in the body at the cellular or molecular level. To reach such a level of detailed imaging requires specialized instruments (modalities) that either alone or in combination with imaging agents (either targeted or passive) can uncover the status of a tissue via detection of biochemical or other molecular markers, particularly in the context of a disease state.
One of the major aims of molecular imaging is to answer questions of disease processes that are beyond current diagnostic methods. The need is to image biochemical processes in intact living organisms over time in a minimally-or non-invasive manner. As such, molecular imaging has the potential to transform each aspect of medical care including improved and earlier diagnosis, more accurate monitoring of therapy, identifying novel drug targets/drug validation and helping to understand the complex biochemical pathways that interact to create disease states.

The requirement for molecular imaging of inflammation in modern medicine
We now know that inflammation is a key or major contributing factor in many chronic diseases including cardiovascular disease (CVD), rheumatoid arthritis (RA), inflammatory bowel disease (IBD), colitis, systemic lupus erythematosus (SLE), diabetes, multiple sclerosis (MS) and Alzheimer's disease and is closely related to many aspects of tumor progression. These conditions contribute a significant proportion of worldwide morbidity and mortality and cause substantial socioeconomic burdens on society (Jacobs, Bissonnette, & Guenther, 2011;Mathers & Loncar, 2006).
Although new therapies have brought significant improvements to patient care, a considerable portion of patients still lack sufficient therapeutic effect and incidence of complications remains high, which is often a by-product of the non-specific nature of the treatments provided. There is also a lack of tailoring of drugs to individual patient needs, which is especially important in chronic pathologies that may present with different underlying causes or have distinct clinical stages, each defined by a specific inflammatory environment. We briefly take RA as an example of where a more sophisticated understanding at the molecular level could improve patient care. RA is a disease that affects 0.5-1% of the global population -treatment options are significantly greater than for many other chronic inflammatory conditions. Approved biological therapies for RA, principally anti-TNF antibodies, were a revolutionary breakthrough in advancing RA treatment, adding to the established broad ranging anti-inflammatory treatments such as methotrexate, yet much is still unknown about the drug-disease interactions such as why some patients respond to anti-TNF more than others or why there is frequently a dissociation between clinical scoring and destructive joint progression (Smolen, Aletaha, & McInnes, 2016). It is also recognized that early diagnosis followed by rapid administration of appropriate therapy is the ideal strategy for treating inflammatory RA and improving patient outlook, yet early diagnosis is challenging due to early symptoms going undiagnosed or not being present at all. Studies of pre-clinical RA include individuals with positive autoantibodies and marginally raised inflammatory markers with no visible clinical features (Firestein & McInnes, 2017). This prodromal period can precede the disease for a decade and poses a significant clinical and therapeutic challenge, which could be transformed by better molecular imaging to guide early clinical decision. Currently, clinicians determine disease presence/severity based on several factors: symptoms, detectable disease biomarkers and medical imaging results. The advent of biological therapies targeting inflammatory molecules such as cytokines can, in turn, also lead to new diagnostic approaches with the addition of imaging agents to the targeted antibody allowing visualization of the cytokine concentration in the disease site and to assess whether the unlabeled antibody could be of benefit therapeutically. Information gained from such an approach may also help answer why certain patients are non-responders to a cytokine-targeted therapy. In comparison with cancer, inflammatory diseases such as RA involve multiple tissues and it is yet unexplored if there is considerable mismatch between disease and treatment distribution in the body. Studies exploring lymph node, bone marrow and synovial tissue biopsies would further support the systemic distribution of the disease (van Baarsen et al., 2013). Hence, the ability to visualize disease potentiating molecules can inform clinicians in diagnosis, treatment monitoring and in deciding whether or not a potential therapy may be beneficial (Malviya et al., 2013).
A final point worth noting is that each disease typically has a specific treatment regime which traditionally has ignored potentially common underlying inflammatory components between different diseases that have a strong immune component. This includes the role of inflammation in cancer (Grivennikov, Greten, & Karin, 2010), and the interrelationships between RA and CVD (Mason & Libby, 2015) or chronic obstructive pulmonary disease (COPD) and stroke (Austin, Crack, Bozinovski, Miller, & Vlahos, 2016). Indeed, adverse cardiovascular events are increased in RA patients and are subsequently reduced following introduction of anti-inflammatory therapies although the inflammatory mediators linking these two diseases are unclear (Tocci et al., 2016). These unanswered questions are a result of a lack of understanding of the precise cellular and molecular inflammatory mechanisms that drive each particular pathology (Tabas & Glass, 2013). Ultimately, inflammatory processes are finely tuned systems, with many molecular signals that orchestrate the immune response. For most conditions, it is probable that only limited diagnostic information will be obtained by one or two blood-derived inflammatory markers. The potential strength of molecular imaging is to 'see' into the core of the inflammatory site, targeting multiple biomarkers (multiplexing) to provide a deeper insight into the triggers, cellular interplay, pro-versus anti-inflammatory mediators and the ability to evaluate current and the next generation of therapies for chronic immune diseases with a level of precision and detail not seen before. Implementing molecular imaging, therefore, can uncover new information on disease processes which can in turn offer new targets for disease diagnosis and stratification. In this sense, molecular imaging could confer a discipline defining transition in the next decade with profound potential to revolutionize the management of patients with chronic diseases.

Overview
The mainstays of current clinical imaging are computed tomography (CT), magnetic resonance imaging (MRI), ultrasound (US) and the nuclear imaging modalities: positron emission tomography (PET)/single photon emission tomography (SPECT). These imaging techniques have transformed medical imaging in improving disease detection, progression and therapeutic assessment across a wide spectrum of diseases from oncology and CVD to rheumatic musculoskeletal diseases and neurodegenerative conditions. In patients, these modalities are primarily used for anatomic and functional imaging at the tissue level, although at least for PET, molecular imaging is being realized, despite significant shortcomings as will be detailed throughout this report. However, they retain inherent weaknesses that preclude the combined high sensitivity, high resolution, and simultaneous detection of multiple biomarkers in a non-invasive manner that is the ultimate goal of molecular imaging research.
Other techniques that potentially offer greater promise as molecular imaging tools include mostly pre-clinical stage technologies such as optical imaging including photoacoustics as well as techniques classified as vibrational spectroscopy which includes Raman spectroscopy and surface-enhanced-Raman spectroscopy (SERS). While pre-clinical technologies such as near-infrared fluorescence (NIRF) with quantum dots and spectral imaging using SERS can allow a high level of multiplexing, these are not yet clinic ready. There exists, therefore, a broad range of established and emerging technologies with molecular imaging potential and these will be described and compared, with examples of use relating to inflammation imaging highlighted. Additionally, for each imaging modality, we offer a section dedicated to potential therapeutic use of the technology in modifying disease processes, for example through using imaging agents loaded with anti-inflammatory drugs. Finally, there will be an assessment of the relative strengths and weaknesses of each technique as a tool for molecular imaging and their prospective use within the clinic.

Imaging agents
Molecular imaging can rely on either direct detection of endogenous molecules or employ exogenous tracers (passive or ligand targeted) that are tailored to a particular modality of choice e.g. fluorescent probes for near-infrared (NIR) optical imaging or radiolabeled tracers for PET/ SPECT imaging. Traditionally, exogenous tracers have been small molecules b2 kDa (b4 nm) in size (e.g. iodinated compounds for CT, chelated gadolinium for MRI and 18 F-FDG for PET). Despite providing significant advances in diagnosing disease, particularly in the field of oncology, small molecule tracers have important limitations with regard to molecular imaging owing to their rapid clearance, lack of adaptation to allow detection via imaging, potential toxicity and relatively poor signal strength. These limitations have led to development of novel materials to overcome the limitations of current diagnostic molecular imaging and to cater for the new technologies that are emerging, particularly within the optical and vibrational imaging fields. Indeed, for several imaging modalities such as magnetic particle imaging (MPI) and SERS, nanocarriers (or nanoparticles (NPs) as they will be referred to throughtout the remainder of this review) are an essential component for signal induction. In comparison to small molecules, NPs offer a high level of control over design (e.g. contrast agent class/dose, body residence times, targeting ligands, drug delivery). NPs are small nanomaterials, less than 1 μM, that can be composed of a diverse range of materials depending on the requirements of the modality and desired residence times. NPs offer the following advantages over small molecules: 1) They can carry a high quantity of tracer to the site of interest (such as bismuth or gold for CT) in a protective shell, creating a concentrated signal to improve sensitivity coupled with reduced toxicity, assuming the elimination kinetics of the NPs are carefully understood; 2) They can be engineered to various specifications to increase circulation/residence times allowing stable longitudinal imaging; 3) Can be used for passive uptake into tissues via the enhanced permeability and retention (EPR) effect or can be functionalized with biomolecules such as antibodies, peptides or aptamers targeted to a disease related biomarker of interest; 4) The large surface area allows many targeting ligands to be attached to a single NP which increases the binding capacity compared with single molecules. Indeed, this increased binding can be quite substantial, ranging from 1 to 2 orders of magnitude to as high as 5-6 orders of magnitude (Benezra et al., 2011;Hong et al., 2007); 5) The adaptability of NPs to a particular modality coupled with an ability to create multiple 'flavors' of NPs (NPs targeted to different in vivo targets) allows high multiplexing potential. The ability to simultaneously measure multiple biomarkers (Table 1) within the same acquisition window is the key to diagnostic molecular imaging, particularly within the context of inflammation imaging when a single inflammatory marker is insufficient to provide a clinician with the information to diagnose/stratify a disease, particularly when pathological changes may be minimal; 6) Can be engineered to be multi-modal (the detection of one NP by two or more imaging modalities); 7) Can offer novel theranostic strategies e.g. by providing a high sensitivity imaging signal coupled with delivering a drug payload.

Hybrid imaging platforms
As observed in the last 2 decades, it is unlikely that one single imaging modality will possess all the required attributes (high spatial resolution, high sensitivity, low toxicity, fast scan times, consistent quantification, deep tissue imaging, strong multiplexing etc.) for clinical molecular imaging. Indeed, with common imaging modalities, there is currently an inverse correlation between sensitivity and spatial resolution. These limitations have led to the engineering of hybrid technologies, many of which are now commonplace such as PET/CT, SPECT/CT and the re-emergence of PET/MRI, all of which are being investigated for molecular imaging potential. Pre-clinically, there is also a push towards multi-modal imaging, such as combining two or more optical imaging technologies onto a single probe, or NIRF combined with intravascular ultrasound (IVUS) or even single multi-modal NPs that are amenable to multiple imaging systems, for example a NP with an iron oxide core functionalized via a dextran coating to fluorophores for optical imaging (Wolfbeis, 2015). This report will describe recent developments in hybrid imaging systems using established imaging modalities in the molecular imaging space and the potential of more early stage pre-clinical multi-modal imaging systems and their route to clinical translation.

Computed tomography (CT)
CT has been a mainstay medical imaging technique since its inception in the 1970s, superseding older X-ray technologies, owing to its greater power in creating 3D anatomical imaging through computational algorithms as opposed to classical 2D X-ray imaging. CT operates with relatively good spatial resolution, at least for anatomical imaging needs (Brenner, 2010). CT can readily distinguish electron-dense structures such as bone and teeth from electron-poor structures such as soft tissues making it an excellent modality for anatomical imaging of skeletal structures. Inherent advantages include fast scan times and body wide imaging. In the absence of a contrast agent, discrimination of anatomical tissues can be challenging (Cormode, Naha, & Fayad, 2014). The most common contrast agents are iodine-based compounds which have found particular use in coronary artery imaging including Coronary Computed Tomography Angiography (CCTA).

Clinical use
CT is primarily an anatomical imaging technique, limited by its low sensitivity, an inability to accurately discriminate soft tissue structures and to accurately quantify contrast agents (Anderson & Butler, 2014). While the advent of dual energy CT (DECT) has greatly improved the technology, clinical advancement has mainly been realized via enhanced detection of lesions, such as in the fields of oncology and CVD. Indeed, CT's main use in CVD has been calcium scoring of plaques to assess atherosclerotic plaque burden and CT calcium scoring is widely practiced for risk stratification within the clinic, although calcification is a product of late-stage pathology limiting its potential in diagnosis of early stage lesions (Sandfort, Lima, & Bluemke, 2015). CT has not progressed clinically to the level of true molecular imaging either in oncology, CVD (the two major drivers of CT technology) or other conditions and the inherent low sensitivity, lack of approved tracers and concerns over toxicity means clinical molecular inflammation imaging is lacking.

Pre-clinical use and in-human trials
Micro-computed tomography or "micro-CT" is similar in principle to standard CT but on a smaller scale and with superior resolution and is frequently employed in pre-clinical research. Micro-CT produces high resolution (typically 15-50 μM in pre-clinical studies) (MI Bioresearch Inc.; Biolaurus Inc.) although some in vivo systems offer better performance still. Micro-CT scanners are employed, both in vivo and ex vivo, typically to measure hard tissue structures. Modifications of micro-CT utilizing ionic iodine contrast have been used to measure glycosaminoglycans (GAGs), a major component of healthy joint cartilage with an inverse relationship existing for joint iodine concentration and cartilage density (Palmer, Guldberg, & Levenston, 2006) and such an approach Early inflammation (pre-clinical symptoms) was detectable without loss of blood-brain barrier integrity van Kasteren et al., 2009 ICAM-1 Targeted MPIOs No (No) Mouse stroke model (pre-clinical) Early inflammation post-stroke was readily visualized in vivo with ICAM-1 targeted MPIOs but not ICAM-1 targeted Gd liposomes. This data supports other studies that MPIOs are superior contrast agents in vivo Deddens et al., 2013 Targeted MPIOs No (No) EAE (pre-clinical) Blood vessel inflammation in the brain was detected by MRI in EAE mice without loss of blood-brain barrier integrity Using a 1-T small animal scanner, atherosclerotic lesions in apoE −/− mice were detected via binding of the tracer to sub-endothelial macrophages within the plaque. Signal was specific to targeted particles and may offer a non-invasive meaure of plaque inflammatory activity with higher signal to noise compared with PET. Woodside et al., 2018 has been used to measure cartilage growth in rats (Xie, Lin, Levenston, & Guldberg, 2009). Multicolor or spectral CT can differentiate several compounds based on each compound's specific energy-dependent Xray absorption profile. One such spectral-CT imaging system (MARS small bore scanner; MARS Bioimaging Ltd) has been used to image an excised human tibial plateau (break of tibia bone at knee joint) with discrimination of different joint components possible in a proof of concept study (Rajendran et al., 2017). Spectral CT potentially overcomes some of the obstacles of clinical level CT and MRI in combining fast, high resolution scanning with more accurate tissue resolution and quantification of contrast enhanced areas but requires further refinements to be a viable clinical tool (Rajendran et al., 2017). Translating this to clinical osteoarthritis use would require utilizing the ionic iodine spectral CT to measure disease progression and/or response to therapy. It is worth noting; however, that radiation dose was not measured in this study and remains a key concern with high resolution CT imaging. Attempts using the MARS spectral-CT system have also been employed with a view to advancing spectral-CT imaging of the atherosclerotic plaque. Using this technology on excised human plaques, spectral-CT was able to differentiate multiple plaque components in human tissue sections within the same voxel including lipid, calcium/iron and water containing components with the caveats that high radiation doses were used and calcium and iron could not be distinguished within the same voxel (Zainon et al., 2012). Cormode et al., 2010 employed gold NPs targeted to macrophages (Au-HDL) together with an iodinated contrast emulsion agent for imaging small vessels in addition to endogenous calcium for spectral-CT in apolipoprotein-E (apoE) −/− mice, the goldstandard model for the study of experimental atherosclerosis. Specific uptake by macrophages in plaque regions was evident and the gold signal could be discriminated from iodine and calcium allowing simultaneous imaging of multiple plaque components (Cormode et al., 2010). Pre-clinical development of NPs as CT imaging agents has also progressed -albeit at a much slower and limited rate compared with other modalities such as MRI and PET -with the aim of targeting inflammation in atherosclerosis. An iodinated NP contrast agent, N1177, was developed with the intention of specifically targeting macrophages and imaging plaque inflammation (Hyafil et al., 2007). Using a rabbit model of atherosclerosis involving double balloon injury combined with a high fat diet, N1177 was used for in vivo detection of macrophage containing lesions using a clinical CT scanner 2 h following intravenous administration. CT images of macrophage density showed good correlation with histological staining (Hyafil et al., 2007). N1177 showed superior imaging of plaque density compared with a conventional CT tracer but the dose of iodinated N1177 employed was high and patient tolerability to a functional dose may well be a limiting factor.

Theranostics
The high atomic number of gold coupled with its low toxicity profile make it a more desirable CT imaging agent than iodine and more likely to gain regulatory approval than some of the more exotic metals beings trialed due to its long history of use in humans. A new avenue emerging in the nanoscience field is the design and synthesis of nanomaterials that offer a therapeutic and diagnostic option with photothermal therapy. The combination of NP design and laser excitation wavelength can induce absorption of light and subsequent conversion to heat. Following NP binding to or uptake by cells, the thermal energy from the NP is transferred via convection to the surrounding cellular medium resulting in thermal cell death. Gold NPs can be shaped to allow absorption of heat from an external light source, creating an excellent inducer of the photothermal effect. CT can be utilized with NIR light sources to measure uptake of gold into cells and subsequent response to thermal destruction of tumor cells (von Maltzahn et al., 2009) and in macrophages within arterial lesions (Qin et al., 2015). As will be outlined in subsequent sections, photothermal therapy utilizing gold NPs is a strategy also being developed for other imaging/theranostic modalities.

Advantages and barriers to translation
The key advantages of CT include speed of imaging, the body-wide view and its established use within the clinic for anatomical and limited functional imaging. As one of the main imaging modalities, CT scans have quick clinical turnover and are well tolerated by a majority of patients. In fact, CT is widely used in the field of inflammation imaging as recently reported by the Antoniades group who demonstrated that changes in the lipid and water content of perivascular fat tissue surrounding human coronary arteries was a marker of the underlying vascular inflammation associated with atherosclerotic plaques and could be detected by CT in human patients (Antonopoulos et al., 2017;Oikonomou et al., 2018). This offers the potential of CT as an early diagnostic tool for indirectly measuring inflammation in the early stages of atherosclerosis. However, the low sensitivity (10 −2 -10 −3 mol/L), radiation burden and a spatial resolution insufficient to discriminate inflammation 'hot-spots' within the relatively large voxel areas (Sandfort et al., 2015) are likely to be insurmountable obstacles for clinical translation of CT as a true molecular imaging modality.

Magnetic resonance imaging (MRI)/magnetic particle imaging (MPI)
MRI is one of the most studied modalities for functional, anatomic and cellular imaging at a clinical level. Non-contrast MRI is readily used for acquiring anatomic information but if molecular imaging is needed, it requires the use of contrast agents owing to poor sensitivity, the main limitation of MRI. While sensitivity of MRI is superior to CT with ranges from~10 μM to~10 mM (Sinharay & Pagel, 2016), it is orders of magnitude lower than that obtained by nuclear or optical imaging systems (James & Gambhir, 2012) and this limits MRI to situations where contrast agents are highly accumulating through passive means or through binding to highly expressed biomarkers in targeted MRI (Table 1).
There are four principal classes of contrast agents used for MRI and each differs in the mechanism of signal generation: 1) T1 and T2/T2* contrast agents T1-weighted MRI consists of positive contrast agents, typically chelates of paramagnetic metal ions, usually gadolinium (Gd 3+ ), with detection limits in the micromolar range. These are generally small molecule chelates with many approved for clinical use. Attempts to overcome the low sensitivity have involved encapsulation of Gd 3+ into various NPs such as dendrimers, micelles or high-density lipoprotein (HDL) type NPs (Kircher & Willmann, 2012). These NPs can allow delivery of larger payloads of imaging agent, improving sensitivity while reducing the likelihood of release of the toxic free Gd 3+ element.
T2-weighted MRI occurs when negative contrast imaging agents are employed. The mainstay imaging agent for T2 MRI has been the superparamagnetic iron oxide particles (SPIOs) owing to their high magnetization properties. These particles are usually dextran coated and have evolved since their inception in the 1980s to smaller more refined coated particles (USPIOs), with a diameter 10-50 nm and these have been the primary focus to date in moving MRI towards a tool more amenable for molecular imaging. There also exists larger iron oxide-based contrast agents called micron-sized iron oxide particles (MPIOs) that are typically greater than 1 μM in diameter. As will be outlined below, however, even those agents to gain regulatory approval have struggled to maintain a foothold in the imaging market.

2) Fluorinated contrast agents
The 19 F nucleus has properties that offer MRI signal intensities approaching that of hydrogen protons (Ruiz-Cabello, Barnett, Bottomley, & Bulte, 2011) yet is present in only trace amounts in the soft tissues of the human body, thus is a suitable choice for an exogenous contrast agent when high signal to noise ratio is required. 19 F contrast agents have centered on perfluorocarbons (PFCs) where the hydrogen atoms are replaced by 19 F atoms and which can subsequently be concentrated/ encapsulated by a lipid bilayer at near nanoscale size delivery vehicles, allowing targeting to molecular markers or drug conjugation for therapeutic delivery. The key advantage of 19 F MRI is that the signal directly correlates to the binding/uptake of 19 F at the site of interest, allowing quantification of the target. Being detectable by classical MRI (if a 19 Ftuned RF coil is equipped), 19 F MRI still requires N10 −5 M of 19 F in the tissue for detection at 1.5-T (Ruiz-Cabello et al., 2011;Terreno & Aime, 2015).
3) Chemical exchange saturation transfer (CEST)/paraCEST CEST and paraCEST compounds utilize a different form of magnetic imaging, namely reducing the magnetization of the water signal with limited effects on the relaxation rate. The process involves part of the CEST agent undergoing magnetic resonance by application of a radiofrequency pulse which promotes saturation of the proton and chemical exchange with water molecules in which the saturated proton is transferred to the water, resulting in saturation of a proton within the water molecule. The remaining unsaturated water molecules can be used to provide an MRI signal (Sinharay & Pagel, 2016). The limitation of this technique is the slow exchange between molecules resulting in poor sensitivity and the requirement for large concentrations (mM range) of target molecules. ParaCEST exclusively employs exogenous paramagnetic agents capable of exchanging protons or water molecules with endogenous tissues. These may include NPs conjugated to paramagnetic ions, thus exposing them to water protons. By tuning the magnetic resonance frequency to the exchangeable saturated protons, contrast signals can be differentiated from the surrounding bulk water. ParaCEST agents offer faster proton exchange and therefore are a more sensitive adaptation of traditional CEST (Viswanathan, Kovacs, Green, Ratnakar, & Sherry, 2010).

4) Hyperpolarized MRI
Hyperpolarized MRI was/is being developed to overcome the poor sensitivity of traditional MRI. Developed in the early part of this century, with first human trials in 2010, it is fast gaining momentum, owing to its potential exponential gains in imaging resolution. The abundance of water in human tissues means only a small fraction of protons in water molecules require magnetic alignment for a signal to be obtained. Thus, MRI has been developed around detection of hydrogen protons. The detection of other nuclei that may be incorporated into an imaging agent such as 13 C or 19 F would be present at concentrations orders of magnitude lower than 1 H and would not be resolved via traditional MRI but can be resolved by hyperpolarized MRI by increasing the % of atomic nuclei in alignment (Golman, Ardenkjaer-Larsen, Petersson, Mansson, & Leunbach, 2003). Several methods exist for creating hyperpolarized MRI agents but dynamic nuclear polarization (DNP) is the most common and can boost signal intensity by 10,000 times  compared with traditional MRI methods.
Magnetic Resonance Spectroscopy (MRS) is a complimentary tool to MRI, used for detecting primarily 1 H atoms within certain metabolites present within a single voxel of interest. To detect a 1 H signal within molecules present at such a low abundance with respect to water and fat, it is necessary to separate (orsuppress) the water signal during MRS acquisition. The acquisition of molecular spectra by MRS can be used to determine metabolic changes that may reflect pathological processes. Since the signals from the molecules of interest in MRS are present at relatively low concentrations, suitable field strengths are required for clinical MRS, with 3-T scanners offering higher sensitivity and the ability to obtain spectra from smaller voxels (van der Graaf, 2010). The particular method of MRS employed depends on what information is required. Typically, MRS uses single voxel spectroscopy (SVS) for a particular localized region of interest. Imaging of several voxels can be used for mapping spatial distribution of a metabolite of interest (e.g across a tumor) and utilizes magnetic resonance spectroscopic imaging (MRSI) but at the expense of quantification accuracy (Faghihi et al., 2017). MRS can identify various molecules within tissue, dependent on concentration and magnetic field strength. These include creatine, choline, N-acetyl aspartate (NAA), alanine, lactate, taurine, glycine, myo-inositol, intracellular lipids and glucose (Verma, Kumar, Verma, Aggarwal, & Ojha, 2016). Of these, NAA, creatine and choline are the most widely utilized and historically, MRS has been a tool applied for studying disorders of the brain.
MPI is a recent technological development (Gleich & Weizenecker, 2005) that directly detects magnetic NPs (almost exclusively iron oxide NPs thus far) and produces positive contrast from the tracer directly (a list of MPI tracers can be found in Panagiotopoulos et al., 2015). MPI offers superior sensitivity and potential gains in spatial resolution over conventional MRI as well as fast imaging times (Gleich & Weizenecker, 2005;Saritas et al., 2013). The basis for these improvements is that the signal (i.e. NP derived) is not present in the body naturally thus offering superior contrast and sensitivity over traditional MRI.

Clinical use
As stated above, MRI is an excellent tool for imaging the soft tissue components within the human body, offering a safer form of imaging, free of ionizing radiation. MRI can form part of the diagnostic work up for investigating abnormalities within the cardiovascular, musculoskeletal, gastrointestinal and other tissues, providing both anatomic and in some circumstances, functional imaging, such as perfusion imaging. The majority of contrast enhanced MRI utilizes T1 based approaches. However, over the history of MRI, a number of iron oxide-based imaging agents have been approved for clinical use, particularly dextran coated NPs, due to low toxicity and straightforward manufacturing (Mulder, Jaffer, Fayad, & Nahrendorf, 2014). Despite a relatively good safety profile (side effects are likely a result of coating material rather than iron oxide per se), use of SPIONs has failed to meet initial expectations and only a select number of agents are still on the market due to poor clinical uptake and low sensitivity/specificity for intended purpose (Smith & Gambhir, 2017).
MRS is a clinically useful tool for studying a range of diseases including brain tumors, Alzheimer's disease and seizures. MRS has also been applied to neuroinflammatory diseases, most notably MS. While inflammatory molecules are not directlty detected, metabolite levels can inform on the severity and clinical stage of the disease, as reviewed in detail by Chang, Munsaka, Kraft-Terry, and Ernst (2013).

Pre-clinical use and in-human trials
Human studies using contrast MRI in the context of measuring inflammation are limited and all have utilized passive uptake of USPIOs/ SPIOs by macrophages. Since these studies depend on T2* contrast changes of iron concentrations within the tissue, either as a result of the EPR effect and/or macrophage uptake, they don't meet the criteria of true molecular imaging. Despite this, the use of SPIONs appears an attractive option for imaging inflammation, particularly in disease states such as atherosclerosis, RA and IBD, where macrophages are a key component of the involved pathologic lesion, allowing either passive uptake via phagocytosis or active targeting to macrophage surface receptors. Readers are referred to reviews by Jin, Lin, Li, and Ai (2014) and Neuwelt et al. (2015) which cover the use of SPIO passive imaging by MRI. To our knowledge, there are no in-human trials currently taking place in the context of inflammation that utilse the other forms of MRI described above. Although several trials involving 19 F-MRI are underway, particularly in relation of cystic fibrosis, the focus of these trials is not on the inflammatory components of the disease. CEST agents based around glucose detection are in early clinical trials for oncology but as of yet, are not designed for measuring inflammation.
There are, of course, a multitude of studies in animal models, utilizing passive SPIOs/USPIOs and 19 F-PFCs to quantify inflammation via macrophage uptake, particularly in the context of atherosclerosis but what follows will only be a review of key pre-clinical work that employ active molecular targeting of MRI compatible imaging agents to keep within the remit of our discussion. Direct targeting of SPIOs via modifications with ligands such as antibodies, aptamers and peptides allows MRI to be used to monitor SPIO binding/uptake at the site of disease, improving signal to noise and diagnostic capability. Key molecules of interest in this area of research have been adhesion molecules including the selectins (P-and E-selectin), vascular cell adhesion molecule-1 (VCAM-1) and intercellular adhesion molecule-1 (ICAM-1), all of which are upregulated on endothelial cells in inflammatory conditions such as atherosclerosis, MS and stroke and are some of the earliest biomarkers signifying cellular inflammation/activation, even before other pathological signs are evident. A summary of targeted MRI studies can be found in Table 1. 19 F-PFCs containing chelated gadolinium targeted to fibrin were first validated by Morawski et al. (2004) in human carotid endarterectomy samples (areas of atherosclerotic plaque) at 4.7-T allowing detection of the thrombus via the targeted localized 19 F signal and associated anatomical details via the T1 signal from gadolinium. Furthermore, the 19 F signal was intense enough to allow spectroscopic quantification of the bound 19 F, a proof of concept that demonstrated that multiple configurations of 19 F-PFCs, each with distinct spectroscopic profiles, could be used to target multiple biomarkers (Morawski et al., 2004). CEST has found research use in detecting enzyme activity, metabolic function (e.g. glucose utilization), hypoxia, neurotransmitter (glutamate) levels, ion concentration and pH (Sinharay & Pagel, 2016). The large concentrations required of either endogenous or exogenous CEST compatible agents would seemingly limit its usefulness in imaging inflammation. Despite this limitation, glutamate, which is amenable to CEST MRI is elevated in inflammatory brain disorders such as MS and amyotrophic lateral sclerosis (ALD). Results have shown that CEST MRI brain images of glutamate (Glu-CEST) distribution align with PET images  at 7-T, showing the potential for CEST to indirectly measure inflammation in neurological disease. Moreover, in animal models of Alzheimer's disease, Glu-CEST can differentiate diseased areas from non-diseased areas (Haris et al., 2013). GAGs in cartilage are also detectable by CEST imaging allowing the tracking of joint cartilage changes in human subjects (Ling, Regatte, Navon, & Jerschow, 2008;Schmitt et al., 2011). ParaCEST imaging agents have been combined with 19 F-PFCs to offer improved sensitivity and quantifiable targeting of NPs. The use of NP sized formulations has greatly increased the potential of paraCEST, specifically the development of lanthanide containing liposomes (lipoCEST) designed to improve the sensitivity of the system down to the nM level (Flament et al., 2013). A key advantage of CEST based MRI is the ability to image two or more imaging tracers in one scan (duplexing), provided the CEST agents display good sensitivity and have sufficiently different resonance frequencies of their mobile protons (Castelli, Terreno, Longo, & Aime, 2013). This has been demonstrated to be feasible for lipoCEST agents in a proof of concept study using local injection of non-targeted contrast agents into the thigh muscle of a mouse with clear discrimination of the two contrast agents visible on the scans (Terreno et al., 2008). LipoCEST agents have also been functionalized with targeting ligands. This was first attempted by Flament et al. (2013) using α v β 3 (integrin present on new forming blood vessels) targeted lipoCEST NPs in a mouse tumor model and although some non-specific binding was observed, it is a starting point towards developing more refined lipoCEST agents which would be desirable given their higher level sensitivity along with the inherent multiplexing capacity of CEST.
The major use of hyperpolarized MRI has been to use 13 C-containing compounds such as pyruvate for imaging metabolic pathways as the 13 C is passed down through metabolites allowing fast imaging of the metabolic state of the target tissue. Recently, the use of [1-13 C] pyruvate imaging in the heart has shown promise for the detection of cardiac inflammation. Lewis et al. (2018) discovered that the metabolism of [1-13 C] pyruvate to [1-13 C] lactate in infarcted murine and porcine hearts was the result of an increased rate of glycolysis in activated monocytes/macrophages, which was linked to IL-1β release. Thus, hyperpolarized MRI may offer a novel method for imaging inflammation following myocardial infarction (Lewis et al., 2018). 13 C-pyruvate has also been researched to detect inflammation through increased metabolism in animal models of arthritis (MacKenzie et al., 2011) and MS (Guglielmetti et al., 2017).
In addition to MS, significant data exists in human studies to demonstrate that MRS can detect metabolic changes indicative of Alzheimer's disease and precursor conditions such as mild cognitive impairment. As the major detectable metabolite in neurons, low NAA can be associated with loss of neuron viability/density while conversely, elevated myo-inositol is a biomarker of glial cell activation/proliferation and these two changes have repeatedly been demonstrated in the brains of Alzheimer's patients (Rami et al., 2007;Zhu et al., 2006). Reduced choline concentration, as an adjunct to NAA and myo-inositol analysis, may also increase sensitivity to brain changes associated with cognitive decline (Tumati, Martens, & Aleman, 2013).

Theranostics
MRI has many features that make it a useful theranostic tool such as good spatial resolution, body wide imaging and a diverse group of combinable carriers and contrast agents that can be adapted to each theranostic application. A recent promising example of iron oxide targeted theranostic NPs was in the context of acute graft rejection. Guo and colleagues utilized CD3 (T cell specific receptor) targeted iron oxide NPs that also contained a therapeutic gene. Both in vitro and in vivo, using a rat heart transplant approach, the authors successfully delivered the gene to T cells, thus reducing the T cell mediated destruction of the transplanted heart with MRI being utilized to monitor the T cell accumulation via the uptaken iron oxide NPs. The effect was shown to be specific to CD3 targeted gene carrying NPs, results confirmed by NIRF imaging (Guo et al., 2012). In the work by Myerson, He, Lanza, Tollefsen, & Wickline, 2011, PFCs targeted to thrombin via an anti-thrombin agent was a novel development whereby the thrombin inhibitor, PPACK, irreversibly binds to thrombin on the clot surface, thus inhibiting thrombin activity (Myerson et al., 2011). Given the low toxicity of PFCs in humans (Palekar, Jallouk, Lanza, Pan, & Wickline, 2015), it is conceivable that targeted PFCs containing 19 F substituted atoms and/or Gd 3+ chelates could be a viable non-invasive tool for delivering and monitoring efficacy of anti-inflammatory drugs targeted to disease specific molecules that show relatively high abundance such as fibrin. CEST and paraCEST agents are being developed and tested preclinically using drug loaded liposomes, to track and monitor drug delivery, mostly in the field of oncology (Castelli, Boffa, Giustetto, Terreno, & Aime, 2014). Proof of concept for monitoring the delivery of antiinflammatories or modulating inflammatory conditions is lacking.

Advantages and barriers to clinical translation
Iron oxide-based agents, having already been approved for human use, have one less barrier to novel uses for in-human imaging. Rather disappointingly, however, several previous iron oxide-based imaging agents have been discontinued due to poor performance or safety issues, with only Ferucarbotran (Resovist®) retaining use as an MRI contrast agent in a few countries. This leads to a requirement for new agents to be developed if T2 MRI molecular imaging is to be adopted more broadly in a clinical setting. Technical issues associated with USPIO molecular imaging that require to be overcome include low signal to noise ratio, poor specificity associated with ability to clearly discriminate signal between tissue locations, long imaging times and the challenge of standardizing conditions between scans due to patient position and magnetic field discrepancies (Ruehm, Corot, Vogt, Kolb, & Debatin, 2001;Sadat et al., 2014;Schmitz et al., 2000). PFCs, which dominate the pre-clinical and early phase 19 F human trials are strong candidates for regulatory approval due to relatively good biocompatibility, biodistribution and favorable toxicity compared with other contrast agents in clinical use. Indeed, PFCs may be safer for human use than metal-based contrast agents (Palekar et al., 2015). Future studies will need to determine if localized 19 F is detectable in human patients at a tolerable administered dose of 19 F-PFC NPs. These prospects will be improved by wider clinical adoption of 7-T MRI which allows higher sensitivity detection of 19 F (Schmieder, Caruthers, Keupp, Wickline, & Lanza, 2015). CEST agents can be utilized in NP formulations to boost sensitivity although most of these studies have only been at the in vitro level. In vivo comparisons of lipoCEST agents with T1 and T2 based liposomes have shown that the CEST signal is more transient due to accumulation in subcellular compartments of macrophages which suppresses the CEST signal. Hyperpolarized MRI was designed to overcome the limitations of 1 H or 19 F based MRI with profound increases in sensitivity possible. At present, only a few molecules incorporating hyperpolarized nuclei such as 13 C have been tested in vivo and have focused on metabolic molecules as markers of tumor progression or inflammation (not dissimilar to the use of 18 F-FDG in PET imaging). The main limitation of hyperpolarized MRI to date has been the transient nature of the hyperpolarized state, making practical use of the technology problematic. Despite this, proof of concept studies have shown that NPs composed of a hyperpolarized 29 Si nuclei can be imaged for over 40 min following injection in mice and with low toxicity (Cassidy, Chan, Ross, Bhattacharya, & Marcus, 2013), indicating the potential for future advances towards wider imaging applications.
MPI is a new technology only developed in 2005, thus current deployments center around limited numbers of small pre-clinical scanners designed for rodent sized imaging (Panagiotopoulos et al., 2015;Saritas et al., 2013). The current state of MPI remains as 'a promising technology' that can take the best aspects of MRI and improve on this with better sensitivity, higher temporal resolution and easier interpretation of scans due to the signal intensity correlating with the iron oxide loaded NPs, without any background signal. Since MPI produces no anatomical detail, it will likely be combined with other modalities with combined MPI/MRI imaging already tested for imaging the heart and vasculature (Kaul et al., 2015). Clinical concentrations of Resovist have shown to be suitable for in vivo MPI imaging (Weizenecker, Gleich, Rahmer, Dahnke, & Borgert, 2009) and the development of new SPIONs tailored to MPI use should greatly increase performance and facilitate the clinical testing of MPI as a highly sensitive, fast, quantitative vascular, cellular and potentially molecular imaging tool (Vaalma et al., 2017).

Nuclear imaging (PET/SPECT)
Like CT and MRI, nuclear imaging offers the ability to image the entire dimensions of the human body. Together, PET and SPECT form the cornerstone of nuclear imaging. PET/SPECT are attractive options for molecular imaging owing to their non-invasive nature and high sensitivity (10 −10 -10 −11 M for SPECT and (10 −11 -10 −12 M for PET) (Stendahl & Sinusas, 2015). High levels of sensitivity allow reductions in the dosage of contrast agents, which is critical when the contrast agent is radioactive or otherwise toxic. At the center of clinical PET imaging is 18 F-FDG, a glucose analogue that is metabolized by metabolically active cells. Tumor cells are highly metabolically active. They undergo rapid rates of division which increases demand for glucose in addition to actively inducing angiogenesis. Increased glycolysis is also a feature of the immune system following activation. This is most evident for those immune cells that can participate in phagocytosis such as macrophages and neutrophils. Therefore, 18 F-FDG can be used to identify areas of macrophage activation/accumulation. 18 F-FDG sits at the boundary of molecular and metabolic imaging, offering high sensitivity detection (although this is frequently associated with poor signal to noise due to non-specific uptake by healthy cells, particularly cardiomyocytes). PET has also seen numerous radiolabeled small molecules being developed that can be targeted to various inflammatory molecules ( Fig. 1) to provide more information on the tissue state than simple energy use. Since PET/SPECT lack anatomical details, to enhance localization and improve diagnostic sensitivity, nuclear medicine modalities are commonly combined with CT in hybrid imaging systems which can provide high resolution anatomic details to be superimposed on the PET/SPECT image. PET/SPECT can also be combined with MRI for similar purposes due to the high soft tissue contrast obtained by MRI.
2.6.1. Clinical use 18 F-FDG is routinely used in the clinic within the field of oncology, both to locate the presence of tumors, disease staging and to assess shrinkage of tumors in response to treatment; however, 18 F-FDG-PET is also increasingly being used for non-oncological purposes in fields covering CVD (Pelletier-Galarneau & Ruddy, 2019), neurology, infection and inflammatory disorders (Zhuang & Codreanu, 2015). Here, we will only summarize clinical use pertaining to inflammatory diseases. 18 F-FDG PET is used for monitoring disease in cardiac sarcoidosis patients (Glaudemans et al., 2013). 18 F-FDG PET is useful where MRI may be contraindicated due to implanted cardiac devices. A further advantage of PET in this context is more accurate diagnosis and the identification of disease in extra-cardiac sites such as the liver, bone and spleen (Patel, Gunasekaran, Goettl, Sweiss, & Lu, 2017) which may be missed by MRI due to a narrow view window. Identifying the disease in other organs other than the heart will allow more accurate staging and to initiate treatment accordingly. Other inflammatory conditions including vasculitis are also imaged in the clinic; this is an increasingly common use of PET but requires the operator to be trained in the specific protocols related to this particular pathology (Hautzel, Sander, Heinzel, Schneider, & Muller, 2008). Due to the poor spatial resolution of PET, imaging of vasculitis is restricted to large caliber arteries such as the aorta but imaging of smaller arteries (usually 1 st order aortic branches) is possible if the camera has sufficient resolution (≤ 2.5 mm) allowing differentiation of inflammatory vascular conditions such as giant cell arteritis (GCA), Takayasu's arteritis (TA), and polyarteritis nodosa (PAN) (Glaudemans et al., 2013). Although some encouraging results have emerged from studies using a small number of patients, there is currently insufficient evidence to justify the routine clinical use of 18 F-FDG in RA, pancreatitis, IBD, tuberculosis and osteomyelitis. In 2013, combined guidelines from the European Association of Nuclear Medicine (EANM) and the Society of Nuclear Medicine and Molecular Imaging (SNMMI) were published for the use of FDG-PET in inflammation and infection, outlining the clinical evidence (or lack off) for the justification of 18 F-FDG in imaging inflammation (Jamar et al., 2013). Readers can refer to this document for a more detailed account of the many issues and criteria that must be addressed/overcome for 18 F-FDG to be recommended as a tool to improve clinical diagnosis. It is also worth noting how few the inflammatory disease states are that the EANM and SNMMI consider having sufficient evidence to warrant clinical 18 F-FDG imaging. For now, at least, 18 F-FDG-PET in the clinic is still primarily a tool for oncologists.

Pre-clinical use and in-human trials
Although 18 F-FDG is employed as a diagnostic tool in inflammatory vasculitis conditions, it is not employed clinically to diagnose atherosclerosis, heart attack or stroke. This is despite atherosclerotic plaques being inadvertently detected in patients undergoing scans for cancer and a number of studies indicating potential clinical value as will now be outlined. A widespread assessment of the likelihood of a person developing CVD is called the Framingham risk score (FRS), originally introduced in 1998 and updated in 2008. Interestingly, an assessment of 1181 asymptomatic subjects administered 18 F-FDG to measure carotid plaque uptake via PET indicated that persons with elevated plaque uptake of 18 F-FDG had an increased chance of an elevated FRS of 10% or more (Odds ratio: 1.9) (Noh et al., 2013). Retrospective studies using 18 F-FDG have also demonstrated that this method independently predicts CVD risk, even beyond that predicted by FRS (Figueroa et al., 2013). Furthermore, 18 F-FDG PET/CT has been employed in human patients to detect symptomatic carotid atherosclerosis (Rudd et al., 2002). One final example where 18 F-FDG-PET can have predictive power in CVD is in predicting recurrent stroke. 18 F-FDG-PET imaging of carotid artery inflammation in prior stroke patients predicts the recurrence of cerebrovascular stroke, independent of other factors (Marnane et al., 2012). These studies and others highlight the potential of 18 F-FDG-PET as a stand-alone tool for assessing atherosclerosis associated inflammation and plaque severity and can yield good signal to noise ratios with little background from healthy arterial walls. This raises hope that by targeting glucose metabolism as an indirect measure of inflammation, PET could also act in co-ordination with classical arterial angiograph methods to guide surgical and pharmacological  intervention. There are several co-morbidities that act to increase the risk of developing CVD. One such disease is psoriasis (Brauchli, Jick, Miret, & Meier, 2009;Prodanovich et al., 2009). Using 18 F-FDG-PET, Naik et al., 2015 demonstrated that psoriasis patients with no history of CVD had increased aortic 18 F-FDG uptake which increased proportionally with the severity of psoriasis. Therefore, 18 F-FDG-PET could be an early diagnostic tool in psoriasis patients for detecting the vascular inflammatory changes that may signify the onset/progression of atherosclerosis. A similar increase in vascular inflammation has also been observed in RA patients, a disease associated with~50% enhanced CVD risk (Rose et al., 2013).
Several studies have employed 18 F-FDG-PET to measure response to anti-inflammatory therapies. To date, there are five classes of drugs used to treat CVD where both clinical endpoint and PET imaging are available. Clinical trial endpoints have shown positive for two classes of treatments (statins and thiazolidinediones) which correlated with 18 F-FDG-PET findings of vascular inflammation (Osborn, Kessinger, Tawakol, & Jaffer, 2017). The ability to measure specific pathogenetically related outcomes rather than awaiting clinically significant CVD events improves time and cost of trials, as well as patient safety in case of ineffective treatments. Therefore, 18 F-FDG-PET can be used to not only detect inflammation in the primary disease but to assess cross-disease risk, monitor disease progression and even monitor response to therapy.
The use of radiolabeled metabolic precursors is not limited to 18 F-FDG. Choline, a molecule involved in the synthesis of cell membranes is taken up by rapidly proliferating cells (such as macrophages during inflammation) and has been utilized for PET as either 18 F-choline or 11 C-choline. A feasibility study in humans showed the potential of imaging vascular changes associated with CVD using 18 F-choline (Bucerius et al., 2008) and this molecule is currently undergoing further evaluation in human trials (ClinicalTrials.gov identifiers: NCT01899014 and NCT02640313).
While Technetium-99 m ( 99m TC) and 18 F-FDG remain central to current nuclear imaging, pre-clinically, a multitude of novel targeted radiolabeled tracers have been developed, many of which target inflammatory molecules (Fig. 1). Table 2 provides a summary of the most important pre-clinical studies utilizing PET and SPECT imaging agents targeted to inflammatory biomolecules. Although, both the clinical and pre-clinical nuclear imaging landscape is dominated by small molecule tracers, NPs conjugated to or containing a core of radiolabeled tracers have been developed. 99m Tc radiolabeled NPs are commonly used in SPECT imaging and usually consist of sulphur colloids ranging from 10 to 1000 nm in diameter. Indeed, with the exception of iron oxide-based NPs used in MRI, 99m Tc labeled colloid NPs remain the only clinically approved NPs for diagnostic imaging (Thakor et al., 2016). Gold NPs conjugated with ICAM-1 targeted antibodies show enhanced gamma ray production in the joints of arthritic rats (Shao et al., 2011). Since ICAM-1 is an early indicator of inflammation in arthritis, this is a key target of interest that could aid early diagnosis in RA in addition to other chronic inflammatory conditions. The first in human testing of a targeted radiolabeled NP has taken place in melanoma patients (Phillips et al., 2014;ClinicalTrials.gov: Identifier: NCT01266096). A more relevant development has been the first receptor-targeted NP agent for cardiovascular imaging approved for clinical trials. The imaging agent ( 64 Cu-25%-CANF-Comb), a 16 nm copolymer based NP that targets natriuretic peptide receptor C (NPRC)-which is upregulated on atherosclerotic plaques-and has completed an early phase 0 biodistribution and safety trial (ClinicalTrials.gov: Identifier: NCT02498379) with a phase 1 trial now ongoing to demonstrate the use of this NP PET agent for identifying atherosclerotic plaques in the carotid arteries via PET/MR (ClinicalTrials.gov: Identifier: NCT02417688). Alternatively, targeted NP based approaches in atherosclerosis are being developed in animal models with CCR5 (chemokine receptor upregulated on atherosclerotic plaques that facilitates immune cell entry into the vessel wall) targeted NPs showing promise in imaging disease in apoE −/− mice (Luehmann et al., 2014).

Theranostics
Direct use of radiolabeled tracers as therapeutic agents in themselves has focused on harnessing the radiolabel to induce cell death in the context of cancer. This involves exchanging the diagnostic gamma or positron emitting radiolabel with a beta emitter. To date, this process has mainly focused on targeting somatostatin receptors (SSTRs) on neuroendocrine tumors where peptide binding to the tumor cells induces internalization of the ligand/radiolabel and accumulation of the betaemitting isotope, leading to destruction of the tumor cells (van Essen et al., 2009). Alpha emitting isotopes incorporated into tumortargeted antibodies have also been tested clinically for a variety of cancers due to their highly destructive localized effects (Dekempeneer et al., 2016). Such approaches are unlikely to achieve such a disease specific cell/inflammatory target that would cause a beneficial therapeutic outcome without negative systemic effects on the immune system as a whole, since disease-specific cells/molecules would not be sufficiently localized for a targeted killing.

Advantages and barriers to clinical translation
PET has revolutionized the ability to non-invasively detect malignancies in the body. Great strides have been taken in terms of hardware and data processing to lower radiation dose, improve resolution and combine the major strength of PET (high sensitivity) with the anatomic resolution available from MRI and in particular CT. Despite these advancements, the development pace of PET in terms of expanding beyond 18 F-FDG and oncology has been slow. Over 95% of all clinical PET scans are 18 F-FDG scans (Goel, England, Chen, & Cai, 2017) and the vast majority are for oncology. Chemistry is not the bottleneck in PET tracer development with a variety of small ligands being radiolabeled and tested to varying degrees in the pre-clinical setting. Indeed, an inspection of the clinical trials database reveals that between 2005 and 2015, there were actually more oncology trials using non 18 F-FDG tracers (n = 288) than using 18 F-FDG (n = 231). The non 18 F-FDG trials tended to be registered later (2010)(2011)(2012)(2013)(2014)(2015), be phase 0-2 and involve a smaller number of participants . This indicates the rapidly expanding area of targeted PET imaging. As outlined above, 18 F-FDG-PET imaging has proof of concept in detecting inflammation in CVD patients, to aid prognostic decision making and even to evaluate novel therapies. An important limitation of 18 F-FDG-PET, however, is its poor performance in reliably detecting 18 F-FDG in coronary arteries, blockages of which are the principal cause of heart attacks. Creation of accurate cardiac images also suffers from motion artifacts and partialvolume effects (Petibon et al., 2017). Moreover, 18 F-FDG-PET is only a gross readout of inflammation, focused on a single parameter, namely energy use, providing no information on the phenotype of the cells from which the increased FDG use is derived from.
For targeted PET beyond FDG, several considerations need attention. The first is the suitability of the tracer. It should possess high potency and specificity for the biomarker of interest to provide good signal to noise and minimize radiation exposure. The use of new radiolabeled tracers also requires on-site facilities with automation of manufacturing, a process which may well be challenging and costly due to complex synthesis protocols. The high costs associated with PET tracer synthesis as well as the installation and maintenance of PET scanners frequently results in PET research being confined to specialized well-funded institutions with a historical focus on a particular area of research, most usually in oncology. These high startup costs and lack of access to novel PET tracers can severely restrict research in novel areas and the progress towards clinical translation. Table 3 summarizes many of the key obstacles that may hamper the translation of PET ligands into the clinic.

Acoustic imaging (contrast-enhanced ultrasound (CEUS))
US is an established medical imaging technology and one of the most used diagnostic modalities in the clinic. US is commonly used for abdominal, testicular, breast, neck and musculoskeletal examinations in its non-invasive form. US can also be applied to assess the status of blood vessels and determine the presence and gross morphology of atherosclerotic plaques and thrombus. Although this can be assessed noninvasively, improved invasive forms of US (IVUS) are being developed. IVUS uses a catheter-based approach for imaging the wall of blood vessels. It's high spatial resolution and ability to discriminate plaque components in a superior manner to invasive angiography has helped transform our understanding of atherosclerosis. IVUS can be thought of as a moderately high resolution (150-250 μm) morphological imaging tool, rather than a true molecular imaging tool. Therefore, a discussion of standard (non-contrast enhanced) IVUS here is not warranted. For true molecular imaging, as per all the techniques described so far, all forms of US require the use of contrast agents (contrast-enhanced ultrasound; CEUS). US contrast agents (UCAs) generate resonance reflection signals, which provide high signal to noise based on increased acoustic scattering properties. The most common contrast agents in US are gas filled microbubbles. Microbubbles have a size of 1-8 μM which confines their location to the circulation, making them excellent UCAs for imaging the vasculature, allowing visualization of micro-vessels normally hidden on standard Doppler US. In brief, microbubbles all share 3 properties: 1) a core containing a compressible and acoustically reflective gas; 2) a protective biocompatible coating (shell) and 3) a diameter of 1-8 μM allowing body wide vasculature perfusion. While microbubbles dominate the CEUS landscape, other nanoscale UCAs have been developed and are currently being investigated in pre-clinical research. These include echogenic liposomes (ELIPs), nanobubbles, nanodroplets and nanoparticles. The use of nanoscale UCAs opens up avenues to target molecules beyond the vessel wall, owing to their smaller size and ability to enter tumors or inflammatory sites through the EPR effect. Although less well tested, they can also be used as vehicles for drug delivery.

Clinical use
A number of microbubble UCAs have been developed and several are commercially available for clinical use and there now exists a long safety record of use in humans (Schinkel, Kaspar, & Staub, 2016). This high safety threshold is due to microbubbles being able to pass into the pulmonary circulation. After injection of microbubbles, imaging can begin after a few minutes. Following acoustic shattering of the microbubbles, the shell is safely removed via the reticuloendothelial system and the gases are released via the lungs (Schinkel et al., 2016). The clinically approved UCAs are all untargeted microbubbles that have found clinical use in improving US resolution of organ abnormalities (Guvener et al., 2017). US has also made dramatic advances in areas of inflammation medicine e.g. in the field of rheumatologyherein US together with power doppler is now used routinely to detect synovial inflammation, tendon damage/inflammation and even to provide preliminary evidence of bone erosion. Furthermore, the use of both non-contrast and CEUS for assessing gastrointestinal (GI) disorders including IBD is increasing within the clinic. US can detect structural changes in the bowel and surrounding mesentery indicative of inflammatory processes while the use of contrast agents enhances diagnostic sensitivity and the distinguishing features of focal inflammation in Crohn's disease (Kucharzik, Kannengiesser, & Petersen, 2017). Although US imaging of the GI tract can be more challenging than other organs due to the presence of gas artifacts, it can be a valuable, safer and more cost-effective method of GI imaging than other clinical modalities (Kucharzik et al., 2017;Muradali & Goldberg, 2015). No clinically approved use currently exists for molecular CEUS imaging.

Pre-clinical use and in-human trials
Although the clinical use of CEUS is yet to reach the level of molecular imaging, the first in-human clinical trial of targeted molecular ultrasound has now been completed (Willmann et al., 2017) and involved microbubbles (1.5 μM diameter, a perfluorobutane/nitrogen core, surrounded by a phospholipid shell) targeted to kinase insert domain receptor (KDR) (Abou-Elkacem, Bachawal, & Willmann, 2015), a key regulator of neoangiogenesis which is expressed on several cancers including ovarian and breast. The imaging agent was well tolerated by Ga-pentixafor arterial uptake was correlated with plaque severity, CVD risk factors and prior adverse CVD events (Fig. 2 40 patients and US molecular imaging showed good correlation with histological analysis of biopsy samples. Importantly, the safety profile of the trial (in which no patients withdrew due to side effects) was similar to that reported for a non-targeted UCA (BR38) (Schneider et al., 2011) indicating that the addition of targeting ligands may not substantially change the in vivo profile of the UCA. Pre-clinical imaging of CEUS in inflammatory conditions has produced some fruitful results with advances in detection, image processing and quantification using protocols based on the destruction-replenishment method and acoustic radiation force (ARF) imaging (Abou-Elkacem et al., 2015). Adhesion molecules have been a keen area of exploration for targeted UCAs in inflammatory disease including IBD and atherosclerosis. We will now briefly consider each of these diseases in turn: US could be employed alongside CT, MRI and endoscopy to offer earlier diagnosis and identify treatment responders from non-responders in the case of IBD where endoscopy is supplemented with medical imaging (particularly CT but also MRI). As an example, microbubbles targeted to adhesion molecules have been used in animal models of colitis, including the use of dual-targeted (P-and E-selectin) UCAs (Wang et al., 2013). This approach of dual-targeting adhesion molecules shows very close correlation with ex vivo histology in terms of grading the extent and stage of inflammation that could justify the clinical noninvasive use of CEUS for staging IBD and quantifying the inflammation present using a molecular imaging modality that offers greater speed and more tolerable contrast agents than CT and MRI. Moving this prospect closer to reality was the study by Wang et al., 2015 using the same clinical grade dual-targeted microbubble as used previously. In a porcine model of ileitis, the authors demonstrated the feasibility of using CEUS to detect intestinal inflammation (Fig. 3). The novelty of this approach was the use of a clinical US scanner on a large animal, an important step towards translating a quantitative molecular imaging approach for inflammatory bowel conditions into the clinic.  VCAM-1 presents early in atherosclerosis, being present on endothelial cells even before the first signs of plaque formation. VCAM-1 targeted microbubbles were able to evaluate the different stages of atherosclerosis in mice. Furthermore, reductions in VCAM-1 intensity could also be used as a marker of efficacy of statin therapy (Kaufmann, Lewis, Xie, Mirza-Mohd, & Lindner, 2007;Khanicheh et al., 2013). Other molecules to have been successfully targeted using microbubbles include ICAM-1 on inflamed endothelium  and plateletendothelial cell interactions using microbubbles targeted against glycoprotein-Ibα (GPIbα) on platelets and von Williebrand factor (VWF) on endothelial cells (Shim et al., 2015), which may help in dissecting the role of platelets in the initiation of atherosclerosis and in the detection of thrombotic events in late stage pathology. In addition to detecting thrombotic events, CEUS may be used to monitor response to treatments that induce reduction in thrombus size/formation (thrombolysis) . Further information on the comparison of US with other modalities for imaging atherothrombosis has been described elsewhere (Wang & Peter, 2017). A recent advance in CEUS signaling employed the ARF technique mentioned above in mice. Not only is this method quantitative, it reduces imaging times substantially. Indeed, the use of VCAM-1 or P-selectin targeted microbubbles in a murine vascular inflammatory model resulted in significant detection of adhesion molecules in the abdominal aorta and an ability to discriminate mice with inflammation from control animals in just 180 s scanning time .
A final example of targeted microbubble US in inflammatory disease comes from myocarditis, an inflammatory heart condition not associated with ischemia. Endomyocardial biopsy is employed for diagnosis yet is an invasive procedure and has relatively poor diagnostic prediction. Microbubbles targeted to CD4+ T cells and inflamed endothelium (two hallmark inflammatory changes in myocarditis) have been used in a mouse model of the disease. In addition, a third microbubble targeted to a generic marker on all immune cells was also employed (Steinl et al., 2016). Using CEUS, the authors could detect increased signal from the targeted microbubbles in animals with moderate disease with signal increasing with disease severity. Furthermore, it was the first demonstration that targeted microbubbles could be used to detect CD4+ T cell infiltration into an inflammatory site (Steinl et al., 2016).
While the transcutaneous non-invasive CEUS studies above used targeted microbubbles as UCAs, most studies utilizing contrast enhanced IVUS have employed targeted perfluocarbon nanoemulsions or liposomes, rather than microbubbles. Studies have included the detection of ICAM-1 (Demos et al., 1999) and tissue factor (Lanza et al., 2000) in porcine atherosclerotic arteries. A more comprehensive study involved the use of liposomes conjugated to 5 markers of inflammation and thrombosis in an atherosclerotic swine model. IVUS imaging was performed at 20 MHz and results correlated well with post-mortem histological analysis. It should be pointed out that this was not a multiplexing experiment. Each liposome preparation had only one targeting ligand and each animal only received one liposome preparation (Hamilton et al., 2004). Nevertheless, these studies suggest that IVUS with targeted UCAs could be used to better detect and characterize vascular injury and atherosclerosis over standard non-contrast enhanced IVUS.

Theranostics
Therapeutic delivery of drugs via UCAs is an attractive option. Theoretically, a 3 μM wide microbubble could carry millions of proteins or small molecules and these carriers display favorable safety profiles, confirmed by their long history of clinical use. The next issues to address after safety are technical, such as ensuring delivery of the UCA to the intended site (such as via molecular targeting or EPR effect) and finally, release of the payload in a manner that is not destructive to the target tissue. The process most investigated for microbubble-based drug delivery is sonoporation. A detailed explanation of this process can be found in Castle et al., 2013. In brief, microbubbles undergo oscillation upon exposure to acoustic energy. At high amplitude exposure, these oscillations expand in intensity and the shell of the microbubble can rupture causing release of the core materials. This process can serve as the release mechanism for the drug. When an oscillating microbubble is in close proximity to a cell (such as bound via molecular targeting or forced near to the vessel wall by the ARF process), transient, reversible increases in the permeability of the cell membrane can occur due to the interaction of acoustic energy and the oscillating microbubble. When the pulse sequence is controlled, cells can undergo reversible permeability without loss of viability (Juffermans, Dijkmans, Musters, Visser, & Kamp, 2006). The use of US and microbubbles for therapeutic drug delivery has already progressed to human case studies and even clinical trials in the fields of cancer (Dimcevski et al., 2016) and for promoting drug delivery through the blood brain barrier (ClinicalTrials.gov Identifier: NCT02343991).
UCAs have also been employed as theranostic agents with a view to utilize them as carriers of thrombolytic drugs to thrombotic areas of the vasculature. In a swine thrombosis model, Culp et al., 2004 demonstrated that microbubbles tagged with the GP IIb/IIIa receptor inhibitor eptifibatide (a thrombolytic drug) caused clot dissolution in 6 of 8 pigs following transcutaneous US while 0 of 7 pigs treated with eptifibatide alone showed clot dissolution. It has been demonstrated that microbubbles passing over an intravascular thrombosis can induce dissolution of the clot when a high mechanical index US pulse is employed (Xie, Lof, Matsunaga, Zutshi, & Porter, 2009). An important development by Wang et al., 2016 involved the use of theranostic microbubbles for simultaneous detection, thrombolysis and monitoring of response via a single dual-conjugated microbubble. The microbubbles were conjugated firstly with the same novel single chain antibody directed against GP IIb/IIIa  along with urokinase plasminogen activator (a thrombolytic agent). These theranostic particles were utilized for US imaging in the carotid arteries of a mouse thrombosis model where they performed a multi-functional role in detecting thrombi, reducing thrombus size and acting as readout of the efficacy of the conjugated urokinase plasminogen activator. When compared with commercial urokinase alone, the microbubbles were just as effective but with a better safety profile owing to greatly reduced bleeding times compared with urokinase alone. This is significant as the increased risk of bleeding is often a contraindication against the use of urokinase and similar thrombolytic agents in many patients with myocardial infarction and stroke. US targeted microbubbles have also been employed in conjunction with liposome carrying drugs to improve local delivery via ultrasound induced cavitation. Sheng et al., 2018 demonstrated that liposomes loaded with basic fibroblast growth factor (bFGF) could be used to increase the delivery of bFGF to the kidney of diabetic rats and that tissue delivery of bFGF was enhanced in the presence of focal microbubble destruction within the kidney (causing greater drug permeability). This resulted in a reduction in inflammation and improvement of renal function. Such a dual approach can not only improve residence times of reagents such as bFGF, which otherwise have short circulating half-lives, but can also increase the uptake of drugs specifically in the target organ (Sheng et al., 2018).
Microbubbles can also be used for the delivery of genes to target tissues (gene therapy). With regard to inflammatory diseases, diabetes has been the main focus of this approach to date with microbubble based approaches being applied to successfully deliver proinsulin genes to the pancreas in rat models (Chen et al., 2006;Chen et al., 2010) with optimized gene delivery inducing almost complete recovery of the β-cell population that persisted for at least 6 months (Chen, Shimoda, Chen, Matsumoto, & Grayburn, 2012), a promising result that offers a safer delivery vehicle of genes to tissue compared with viral based methods. Successful use of US targeted microbubbles has also been demonstrated to reduce clinical scores in a rat model of arthritis with microbubbles providing local delivery of plasmids encoding the tumor necrosis factor (TNF)α receptor (TNFR) gene (Wang, Tang, Xiang, Tang, & Qiu, 2019). IVUS has also been deployed for therapeutic purposes both ex vivo and in vivo. ARF uses the energy/momentum of acoustic waves to propagate particles in the directional flow of the energy (Dayton, Allen, & Ferrara, 2002) and can therefore be used to direct drug delivery to specific areas of pathology. When combined with microbubbles, an IVUS catheter with a diameter compatible with human coronary arteries was found to function in a model system involving delivery of a fluorescent small molecule to swine arteries ex vivo (Kilroy, Patil, Rychak, & Hossack, 2014) and in vivo (Kilroy, Klibanov, Wamhoff, Bowles, & Hossack, 2014). Importantly, fluorophore delivery was localized to the area where the ARF pulse was employed (Kilroy, Klibanov, et al., 2014) thereby demonstrating the feasibility of guided delivery of drugs to focal areas of the artery wall. A similar device was used in vivo in a swine following vascular injury induced by balloon angioplasty. Arteries treated with rapamycin (immunosuppressant drug) loaded microbubbles and subjected to 5.1 MHz IVUS displayed a 50% reduction in neointimal formation when compared with arteries treated with rapamycin loaded microbubbles alone (Kilroy et al., 2015).

Advantages and barriers to clinical translation
US has a number of advantages that may allow expedited advance into the clinic as a molecular imaging tool. It is a relatively simple, well-tolerated and safe method of imaging that is well accepted in routine clinical practice by both patients and clinicians. In comparison with other routinely used imaging modalities such as CT or MRI requiring radiologist review, specialized US imaging is typically performed at the bedside by specialists requesting the investigation (thus minimizing delays, interspecialty variability and chance of miscommunication). The benefits of US use, therefore are many and can be summarized as follows: 1) high spatial and temporal resolution; 2) no ionizing radiation; 3) real-time data analysis and processing; 4) ease of operation; 5) portable; 6) relatively low cost; 7) long established use in clinic; 8) few side effects; 9) ability to acquire anatomic and molecular information in a single scan; 10) good potential for quantitative data; 11) attractive to the clinician who is a non expert and in private practice. The short interval between contrast agent injection and commencement of imaging is a distinguishing feature of US from other modalities. Microbubbles have a very short half-life in vivo (3.5 mins; Willmann et al., 2008). The key advantage to this is that imaging of the bound targeted UCA can begin shortly after injection (several minutes), unlike PET or SPECT which usually require at least 1 h between administration of the tracer and commencement of imaging (Kircher & Willmann, 2012). This feature of CEUS also allows repeated scans or additional scans of other anatomical areas during the same imaging session.
Promising pre-clinical work on large animals is driving quantitative guided CEUS forward at pace; however, more advanced algorithms will be required for human patients to account for variations in US frequency and imaging depths in addition to more advanced motion-correction methods to account for motion artifacts (Abou-Elkacem et al., 2015;. A limitation of quantitative CEUS is that it may preclude the real-time imaging that is such a strong advantage of current clinical US; however, this area is constantly evolving with ever more refined systems being developed. If US is to be considered a viable alternative to nuclear imaging in clinical practice, these techniques will require clinical grade systems with improved quantitative volumetric analysis including three-and four-dimensional imaging in the presence of UCAs (Castle et al., 2013). A recent technical review on the state of three-dimensional real-time US imaging may be of interest to some readers (Huang & Zeng, 2017). The examples presented above in relation to guiding and localized release of a drug is a major asset for any theranostic modality in achieving therapeutic levels of a drug at a target site without potential systemic complications. Therefore, utilizing UCAs with site directed US to induce microbubble release only at the desired location is of clinical value.
One significant disadvantage of CEUS compared with optical and vibrational modalities is a lack of multiplexing potential within a single scan. This could severely restrict the ability of UCAs to be used to stratify inflammatory diseases and make diagnostic decisions. The sequential use of microbubbles targeted to more than one inflammatory molecule can provide some extra value in terms of enhanced specificity and information but the inability to separate the signal of multiple targeted UCAs in a single imaging session is a clear limitation. Another disadvantage already touched upon is that the current CEUS imaging landscape is dominated by microbubbles, which restricts the disease biomarkers to the endothelium thus reducing the spectrum of inflammatory targets available, such as those present on sub-endothelial inflammatory cell infiltrates.
With the first clinical trial for molecular targeted CEUS completed successfully (Willmann et al., 2017), this should lead to further and more expansive trials in oncology, which will hopefully increase the pace of more sophisticated and accurate quantitative software and more robust standardization of treatment regimens that could readily transfer to clinical studies looking at inflammatory markers. With fewer safety barriers to regulatory authorities, coupled with cost effectiveness compared with nuclear imaging, CEUS is well placed as both a diagnostic and theranostic molecular imaging tool. Indeed, the studies outlined in this chapter that demonstrate UCAs can be efficient carriers of immune-modulating drugs and genes paves the way for an exciting future for US as both a site-specific delivery modality combined with real-time monitoring of therapeutic outcomes.

Fluorescence imaging
Fluorescence imaging involves the detection of emission spectrum from a fluorescent molecular probe that has been excited by a laser at an appropriate wavelength. Probes can be targeted or not and can act as 'smart probes' which are molecules that are conditioned to be activated at a particular site within the body, for example via enzymatic conversion from a non-fluorescent substrate to a fluorescent product. In addition to a 'yes' or 'no' response as to whether a molecule is present, the intensity of the fluorescence produced can be used to measure the activity of the catalytic enzyme, thus potentially signifying a shift in disease state (Scales et al., 2016;Wildgruber, Swirski, & Zernecke, 2013). Fluorescent imaging typically utilizes the near-infrared spectrum (wavelengths between 650 and 1000 nm) and is thus referred to as near-infrared fluorescence (NIRF) imaging . The advantages of imaging at these wavelengths are the lack of photonic absorbance from biological components such as hemoglobin, lipids and water, allowing greater depth of imaging. There is also less tissue autofluorescence at NIR wavelengths, thereby increasing signal to noise ratio and hence specificity (Suter et al., 2011). NIRF imaging has a number of advantages including good spatial resolution (rapidly decreasing with depth) and sensitivity, relative low cost, lack of ionizing radiation and importantly for disease detection, the ability to discriminate multiple fluorescent signals due to discrete emission spectrums. NIRF imaging can be performed ex vivo on tissue sections and in vivo, either noninvasively or more commonly via intravital microscopy (IVM) or through the use of catheter/probe systems. Improvements in NIRF imaging probes are advancing at pace with significant gains in brightness, solubility and stability as well as the range of fluorophores available (Hilderbrand & Weissleder, 2010). Fluorophores are typically either proteins (e.g. green fluorescent protein (GFP)), small molecules or quantum dots (QDs). QDs are small (b10 nm) fluorescent NPs typically consisting of a semiconductor metal core encapsulated in a shell (usually ZnS or ZnSe) amenable to functionalization and fine tuning of particle properties. This ability to modify QDs to improve their properties has made them the most valuable imaging tool in NIRF microscopy, having increased brightness and narrower emission spectrums than other fluorescent based probes (Stevenson, Hueber, Hutton, McInnes, & Graham, 2011).

Photoacoustic imaging (PAI)
PAI, also referred to as optoacoustics, is an imaging modality that uses nonionizing optical radiation and ultrasonic detection. Short laser pulses (typically nanosecond long light pulses) directed at tissue creates US waves due to transient thermoelastic expansion that can be detected non-invasively by a detector outside the patient. Recent advances in both laser technology and detection strategies have seen a significant advance in the capability of PAI. By combining the advantages of both optical and acoustic imaging, PAI can overcome the main limitation of optical technology, namely restricted imaging depths which precludes non-invasive imaging of any sub-dermal human tissue. PAI can image in model systems depths in the 5-6 cm range and superficial structures can be imaged at a spatial resolution of~5 μM, superior to that obtainable from standalone optical microscopy . Like fluorescence imaging, there is a trade off in PAI between resolution and depth with spatial resolution diminishing at deeper depths due to energy dissipation/dispersion of the emitted photons through tissue. Certain endogenous molecules can be detectable by PAI including melanin and hemoglobin, but these are relatively weak absorbers of light and diagnostic use is restricted to a narrow range of physiological functions or pathologies such as vascular perfusion or skin cancer diagnosis. To increase the intensity and scope of PAI requires the use of exogenous contrast agents of which strong light absorbing NPs are the most utilized. NPs absorb light from short applied laser pulses, resulting in heating, thermal expansion and the emission of broadband US waves that can be detected across a range of frequencies by an US transducer. NPs have been utilized in PAI for over 10 years, but recent developments have led to a 100-fold increase in the detection level of contrast agent, bringing photoacoustic sensitivity down to the picomolar range (de la Zerda et al., 2012). A versatile range of nanoscale materials have been tested as PAI contrast agents including both organic (melanin, carbon and semiconducting polymer based) and inorganic (gold, TiS 2 ) based materials. Detailed descriptions of the current probe designs used for photoacoustic/MSOT imaging can be found in reviews by Jiang &Pu, 2017 andGujrati, Mishra, &.

Multispectral optoacoustic tomography (MSOT)
PAI has a powerful ability to spatially resolve optical energy absorption signatures via excitation at a single wavelength. The next step towards a more powerful molecular imaging tool is the use of multiple excitation wavelengths to identify multiple endogenous or exogenous contrast agents (multiplexing). This method exploits the different absorption properties of the target tracers/molecules. The multiplexing capacity of MSOT creates a potentially powerful imaging tool where endogenous molecules or external tracers can be identified based on their absorption spectra. While MSOT can be used to differentiate endogenous photoacoustic detectable molecules, a potentially more important development is the capacity for targeted molecular imaging. This requires the use of tracers/nanomaterials with steep absorption curves (or spectral bands) that can be resolved within a single imaging session through the separation of narrow spectral bands.

Clinical use
None of the optical techniques just described are used clinically in relation to inflammatory disorders. Indeed, only NIRF imaging is used at all on a clinical basis. To date, the only clinically approved fluorescent contrast agents are fluorescein, methylene blue and indocyanine green (ICG) with ICG being the only compound that is actively used for fluorescence imaging (principally for lymph node mapping and solid tumor detection). Optical coherence tomography (OCT) is a highresolution morphological imaging tool that is increasingly commonly used clinically for measuring the status of blood vessels. OCT offers superior resolution to IVUS when phenotyping atherosclerotic plaques, being able to measure the thickness of the fibrous cap in addition to providing other more detailed morphological information (Fenning & Wilensky, 2014). Despite OCT being a valuable tool in diagnosing and investigating atherosclerosis, the Providing Regional Observations to Study Predictors of Events in the Coronary Tree (PROSPECT) trial indicated that morphological information alone was not sufficient to predict subsequent coronary plaque ruptures (Stone et al., 2011) thus highlighting the need for molecular imaging modalities to supplement established technologies. Despite OCT not bring a true molecular imaging tool its inclusion here is due to its high spatial resolution to pin point regions of interest that when combined with a molecular imaging technique such as NIRF, can create powerful multi-modal imaging devices with examples listed below.

Pre-clinical use and in-human trials
Out of the modalities listed above, NIRF has the most pre-clinical studies in chronic inflammatory disease (almost exclusively CVD at a molecular level). These studies vary in design and complexity, utilizing a variety of methods, both invasive and non-invasive to harness the strengths of NIRF imaging and overcome the major limitations.
Fluorescence molecular tomography (FMT), a volumetric imaging technique capable of highly sensitive (picomolar level) 3-D detection of fluorescent probes, can improve localization and deeper tissue imaging (Ntziachristos, Tung, Bremer, & Weissleder, 2002). FMT allows simultaneous quantitative molecular imaging of one or more tracers labeled with dyes operating with distinct NIR spectra and has been employed in pre-clinical studies on rodents. Table 4 lists the key studies in the cardiovascular field of NIRF imaging that utilize exogenous tracers with both non-invasive and invasive studies included. In addition to the studies in Table 4 that employ exogenous tracers, some studies have also taken advantage of the naturally occurring NIR fluorescent molecules that exist in the body to perform NIR autofluorescence (NIRAF) imaging without any external tracer. NIRAF has yielded interest in the fields of cancer, ophthalmology and CVD. Combined OCT-NIRAF plaque imaging using an intravascular optical fiber catheter has been used in patients undergoing percutaneous coronary intervention where NIRAF signal intensity correlated with plaque stage and high macrophage density regions as assessed by OCT . A later study employing a vulnerable plaque mouse model identified the blood breakdown product bilirubin as the source of the autofluorescence in that model, with bilirubin resulting from intraplaque hemorrhage (Htun et al., 2017). This is an important finding since intraplaque hemorrhage is a key histological feature of human plaque instability (Virmani et al., 2005). Combining these data with the patient data by Ughi et al., 2016 suggests that in humans bilirubin may also be the source of the NIRAF; however, other molecules such as oxidized lipoproteins may also contribute and further work will be required to optimize the detection wavelength and define in more detail which autofluorescent molecules are present and at what stage of pathology in human atherosclerosis. The few studies to date suggest a tantalizing prospect for intravascular NIRAF as a modality to reliably detect the vulnerable plaque.
PAI, while combining strong attributes from both optical and acoustic imaging has shown a slow development, with few imaging agents tailored to molecular PAI (Gargiulo, Gramanzini, & Mancini, 2016). Early studies using gold-based NPs demonstrated that PAI could reliably detect adhesion molecules in cells and isolated atherosclerotic vessels (Wang et al., 2009). Gold nanoshells conjugated to VCAM-1 are also able to bind endothelial VCAM-1 and be detected by PAI while displaying no toxicity in mice (Rouleau et al., 2013). Fluorescence microscopy struggles to provide accurate spatial localization which hinders quantification due to scattering effects of the photons within tissue. To attempt to localize with high precision activated proteases, Razansky et al. (2012) employed MSOT with a protease activatable probe and subsequently measured the absorption changes of the activated probe. The authors were able to measure with high specificity and spatial location the presence of the activated probe in isolated atherosclerotic human arteries. The added benefit of optoacoustics is that it reveals the underlying morphology of the plaque, allowing localization of the protease in 3 dimensions (Razansky et al., 2012).
The fact that the spatial resolution of PAI/MSOT is not affected by scatter is a significant advantage over comparative fluorescent approaches. Perhaps the most interesting animal study to date utilized an ICG conjugated PEGylated polymer in a novel nanoprobe (ICG@ PEG-Ag 2 S) for imaging in mice using an inVision 256 PAI system (iThera Medical). ICG@PEG-Ag 2 S was designed for long residence times and being lipophilic in nature, should show preferential homing to atherosclerotic plaques. Interestingly, following administration of ICG@PEG-Ag 2 S, excellent aortic localization was demonstrated with signal lasting for up to 24 h. This was in contrast to mice injected with free ICG where signal was 6-fold lower and had dissipated by 1 h. The authors, therefore, had successfully tested a high contrast real-time PAI system in an atherosclerotic model (Wu, Zhang, Li, Li, & Wang, 2016). MSOT has also been employed to assess adhesion molecule expression in animal models of RA using fluorescently targeted nanocarriers. Using an Lselectin/P-selectin targeted probe in arthritic mice, Beziere et al. (2014) were able to non-invasively image the arthritic joint at a depth of~5 mm, superior to what could be achieved by fluorescence. An imblanace in redox signaling, resulting in enhanced reactive oxidant species (ROS) production, is a central component of inflammation, including vascular inflammation. Recently, an elegant study employed a dual glutathione (GSH)/hydrogen peroxide (H 2 O 2 ) probe with differential absorbance wavelengths allowing a measure of redox status. Use of this probe provided information on both oxidant and anti-oxidant levels within the plaque of LDLR −/− and apoE −/− mice, a result attributed to changes in macrophage number/activation. Importantly, the information gained from redox status could be used to differentiate stable from unstable plaques (Gao et al., 2018). Finally, MSOT has also been used to detect endogenous signs of inflammation (predominately oxygenated hemoglobin) in the colon in a mouse model of colitis using non-invasive imaging with a resolution of 75 μM. The non-invasive imaging results by MSOT correlated well with the diagnostic data obtained from colonoscopy (Bhutiani et al., 2017). Turning to human studies, since PAI/MSOT can image both oxy and deoxy-hemoglobin (allowing oxygen levels and angiogenesis to be assessed) without external contrast agents, PAI imaging of joints in human RA patients has revealed the potential of PAI for early non-invasive diagnosis of RA. Jo et al. (2017) utilized a novel point-of-care PAI system to measure the oxygenation of hemoglobin in the joint tissue of RA patients and healthy subjects. Hypoxia, which is a signature feature of the inflamed joint in RA, related to inflammatory cell infiltration and oxidative stress, was found to be visualized in the joints of RA patients whereas healthy subjects showed significantly higher oxygenated hemoglobin (normoxia) (Fig. 4). This promising result in 10 patients is currently being expanded upon in a clinical trial of 120 RA patients (ClinicalTrials.gov Identifier: NCT00748254). Following on from promising results in mice utilizing PAI for imaging the inflamed colon, a pilot study is currently underway First tri-modal NP tested in vivo for detecting macrophages in apoE −/− mice. Each modality showed good inter-mode correlation with doses below those used clinically in PET studies Nahrendorf et al., 2008 Cysteine-protease ProSense750 (Cysteine-protease activatable probe) Employed a NIRF probe on a clinical grade catheter to measure protease activity in the aorta of atherosclerotic rabbits in real-time following injection of protease activatable probe Jaffer et al., 2008 Cysteine-protease Prosense VM110 (Cysteine-protease activatable probe) Similar study type to that performed above by Jaffer et al. (2008). Important advances towards clinical translation include: 1) a NIRF probe capable of imaging 360°with accurate geometry and co-registration with IVUS 2) Able to image larger caliber (human scale coronary) arteries in blood Key finding is that an FDA approved imaging agent can detect plaques in a rabbit CVD model at a dose/kg that is safe for humans Passive uptake by atherosclerotic plaque ICG Yes (Yes) Atherosclerosis (pre-clinical) Tested a dual OCT-NIRF intravascular catheter with ICG contrast agent in swine models of atherosclerosis and arterial injury. While NIRF could assess plaque severity and inflammation, ICG showed lack of specificity in plaque (due to its lack of fine targeting) Successfully imaged inflamed joints in a mouse model of arthritis and modelled potential clinical imaging depths using simulations. Bhatnagar et al., 2019 to investigate the usefulness of MSOT in diagnosing inflammatory bowel disease (ClinicalTrials.gov Identifier: NCT02622139).
2.8.6. Theranostics PAI/MSOT is perhaps the only optical imaging technology discussed that has a potential therapeutic use due to its extensive utilization of metallic nanoscale NPs and its inherent reliance on the induction of heat in the photoacoustic target. Under laser irradiation, PAI contrast agents produce heat which can cause thermal ablation and death of targeted cells, a process termed photothermal therapy. Hyperthermic approaches can also be used to induce drug release from a nanocarrier at the intended release site (Fomina, Sankaranarayanan, & Almutairi, 2012), with the avoidance of potential systemic side effects and concentrating the drug locally. In an in vitro thrombus model, the acoustic pulse pressure produced under laser excitation creates a cavitation bubble which could promote local drug release/delivery into the vessel wall (Shangguan, Casperson, Shearin, & Prahl, 1996). A more recent and potentially elegant multi-modal delivery system was created and takes advantage of the fact that the photoacoustic effect is not just generated by thermal expansion but also by vaporization. The system involves a protein shell carrier loaded with PFC nanodroplets and optically absorbent gold nanostructures. Under optical excitation, the heat generated by the gold nanostructures induce a liquid to gas phase transition (vaporization) of the PFCs, resulting in a strong photoacoustic signal. Since the photoacoustic effect is greater through vaporization than thermal expansion, this system offers the potential for more powerful PAI as well as a tracer capable of acting as a targeted drug delivery vehicle (Wilson, Homan, & Emelianov, 2012). In fact, this process has already been validated in an in vivo cancer model using the anti-cancer drug, paclitaxel, in combination with tumor cell targeted photoacoustic NPs, with the result being local drug release and tumor destruction (Zhong, Yang, Wen, & Xing, 2016). Other developments include even more complex multi-modal NPs for drug delivery and release, the most interesting being a fluorescent/PAI/CT compatible NP for photothermal therapy and/or drug/gene delivery and such a NP has recently been successfully tested in a murine cancer model (Duan, Yang, Zhang, Zhao, & Xu, 2017). More recently, PAI has been used in combination with NIR light to both monitor retention of dexamethasome loaded nanocarriers and to induce photothermal release of the drug following injection of the nanocarrier into the joint cavity of mice with osteoarthritis. Such an approach allows a targeted high dose administration of an antiinflammatory compound into the joint with only minimal systemic effects (Zhao et al., 2019).

Advantages and barriers to clinical translation
Optical imaging, encompassing all its forms from microscopy to endoscopy and specimen/biopsy analysis is the most common clinical imaging tool. Yet, as has been outlined throughout this chapter, their use for in vivo imaging is severely restricted by poor penetration depths accompanied by decreasing spatial resolutions, lack of biocompatible fluorescent probes and challenges in quantification. All these issues severely limit the potential for clinical translation. Several methods are being developed in an attempt to improve NIRF imaging depths while also allowing quantification. FMT-CT has evolved as a tool for quantifying with greater precision the anatomical location of fluorescent imaging agents. While FMT-CT offers a high sensitivity tool for quantitative assessment of a fluorescent signal within a region of interest, it doesn't overcome the poor spatial resolution of fluorescence at deeper imaging depths, hence only large changes in pathology can be discerned and it becomes impossible to localize the signal precisely without doing subsequent ex vivo correlation experiments on isolated vessels or tissue sections thus negating any benefit of non-invasive imaging.
Since depth of imaging is not of such high importance in intravascular NIRF, it is much more feasible that imaging of arteries -with all the added molecular detail that NIRF can bring, especially when combined with OCT-will be the first clinical use of NIRF in assessing an inflammatory disease (atherosclerosis). Before this can happen though, it needs to be established in controlled trials that NIRF can demonstrate better risk stratification than current methods such as coronary artery calcium scoring and FRS.
PAI/MSOT is now a fast-moving technology with significant potential in the field of medical imaging. It can offer higher spatial resolution than NIRF microscopy, can provide anatomic and functional information in the absence of external contrast agents or additional hybrid modalities, which can be extended to molecular imaging by targeted external contrast agents such as metallic NP based tracers. Furthermore, MSOT extends PAI to a multiplexing modality by being able to distinguish different spectral signatures and therefore assessing different molecular pathways or events (Gujrati et al., 2017). It is clear that PAI differentiates itself from other optical modalities by offering high resolution imaging with mm to cm penetration depths. PAI utilizing NIR light can move beyond NIRF imaging (depths~0.5-1 mm where spatial resolutions can be sub 20 μM) (Zhang, Maslov, Stoica, & Wang, 2006) extending out to several cms while maintaining resolutions in the 100 μM range (Ma, Taruttis, Ntziachristos, & Razansky, 2009). Entire body imaging is possible on small rodents, in particular albino mice, which lack melanocytes whereas the more pigmented skin in, for example, C57BL/6 mice may result in interference from superficial melanin when imaging deeper tissues. In humans, larger organs including human breast (Ermilov et al., 2009) have also been visualized. Since PAI receives acoustic signals, it shares the advantage of real-time imaging with US, allowing dynamic processes to be viewed in living organisms. The advantages of PAI and in particular MSOT, in addition to real-time imaging and improved spatial resolution, are versatile contrast agents, portability, scalability and cost effectiveness coupled with a lack of ionizing radiation (Ntziachristos & Razansky, 2010). Like NIRF imaging, the major limitation regarding human use of PAI is penetration (sensitivity of detection at increasing depth). While the spatial resolution of PAI has been studied in some detail, less studied is the sensitivity of detection of contrast agents. In vivo estimates vary from study to study ranging from sub nM (de la Zerda et al., 2012(de la Zerda et al., , 2010 to low nM detection (Haedicke et al., 2017). Variables that effect sensitivity include depth, type of contrast agent, magnitude of light applied, and the US detector employed. Like NIRF, in terms of PAI being used for in human imaging of inflammatory diseases, this may initially involve intravascular (CVD) or endoscopic (IBD) devices where the restricted depth and other tissue penetration limitations are not a limiting factor. In these circumstances, the high resolution, multiplexing capability of MSOT could bring significant clinical value over current techniques such as OCT or IVUS in identifying cell types or inflammatory signatures within the tissue, thus reflecting the stage and severity of the pathology. Additionally, PAI has been demonstrated to be capable of imaging human joints, potentially bringing arthritis into its sphere of clinical utility (Wang, Chamberland, & Jamadar, 2007). One last area that requires extensive research is quantification of the photoacoustic signal. MSOT, in principle can identify multiple biomarkers by their absorption spectra. However, there are challenges associated with multi-wavelength measurements arriving through biological tissue. Tissue attenuation of light is directly related to depth and wavelength and requires sophisticated algorithms to retrieve and quantify the signal, a feature made more complex when multi-spectral analysis is required. However, significant progress is being made in improving the software data processing of PAI and indeed, NIRF imaging. The goal then is the setting of common standards between instruments and institutions that would be required for clinical level use, alongside stable laser setups and ease of use hardware and software.
Clinical trials in PAI are now underway (predominately first inhuman or phase I) and these will focus on feasibility, safety and proof of concept effectiveness. While many of these trials are in oncology (as described by Valluru, Wilson, & Willmann, 2016), most interestingly for the reader are the two trials outlined above in respect of RA and IBD. Early results from RA are promising as described by Jo et al. (2017). Optimism too is warranted with the first results to emerge from the clinical trial in IBD where MSOT analysis of hemoglobin in the colon wall could distinguish between active Crohn's disease and patients in remission (Knieling et al., 2017). If the outcome of these trials and other phase 1 trials are successful, then larger phase II trials should follow to demonstrate value and reimbursement pathways for regulatory bodies that will determine clinical translation potential assuming biological effectiveness and safety meet the endpoint criteria.

Raman spectroscopy/surface-enhanced Raman spectroscopy (SERS)
Raman spectroscopy is a form of vibrational spectroscopy that can provide data through molecularly specific spectra which are based on the vibrational energies of chemical bonds within molecules following excitation with a light (laser) source. This approach can provide detailed information on the molecular composition of cells and tissue. Raman spectroscopy can also identify multiple molecules in solution or within tissue through a high multiplexing capacity, allowing dynamic changes of molecular composition to be followed over time. The spectra produced by biological molecules creates a chemical fingerprint of the sample allowing detailed analysis of the changing molecular phenotype of a sample at a spatial resolution down to 1 μM (Bunaciu, Hoang, & Aboul-Enein, 2017). Classical Raman spectroscopy is limited by low depth (μM range) imaging and therefore its use will be confined to ex vivo analysis of blood/urine and tissue biopsies or for analysis of skin disorders or placement within modified catheters for invasive use; however, advances in Raman technology now allow for deeper imaging, making in vivo non-invasive imaging a realistic prospect as will be described below.
Raman spectroscopy is highly capable at characterizing the biochemical changes within cells and tissue. Raman generates fingerprint spectra within the 2000-400 cm −1 wavelength window. Raman obtains spectra from the inelastic scattering of light (visible, NIR or UV) from a laser. This is when incident (or arriving) photons from a laser interact with molecular vibrations, resulting in the emitted photons having a different wavelength and these wavelength shifted photons (inelastically scattered) can be used to create molecularly specific fingerprint spectra. Molecular distribution maps of proteins and lipids can be created by plotting the signal intensity of a peak of interest at each pixel imaged. For imaging beyond superficial tissue, a modified form of Raman called spatially offset Raman spectroscopy (SORS) may be deployed. These attributes make Raman an attractive tool for non-destructively acquiring biological information; however, it does have a major disadvantage, namely the inherent weakness of the Raman effect. Only about 1 photon in a million undergoes inelastic scattering resulting in poor sensitivity which results in long imaging times and an inability to discriminate low-level target analytes from background noise. Finally, many biological molecules are spectrally similar requiring complex chemometric analysis to separate spectra of interest . To overcome the low sensitivity of Raman, two modifications to Raman spectroscopy, termed stimulated Raman scattering (SRS) and coherent anti-Stokes Raman scattering (CARS) have been developed. Classical Raman is often referred to as spontaneous Raman spectroscopy when used in comparison to these two other modifications. Both SRS and CARS are orders of magnitude more sensitive than spontaneous Raman, utilize more than one light source and depend on specific resonance parameters between the lasers and the molecules being analyzed. A detailed explanation of SRS and CARS can be found in a well-written review by Tipping, Lee, Serrels, Brunton, and Hulme (2016). Both SRS and CARS allow rapid image acquisition, a major advantage over spontaneous Raman, although they probe only one vibrational signature at a time as there are no spectra obtained from the non-resonant background (Laing, Jamieson, Faulds, & Graham, 2017).
The shortcomings associated with Raman spectroscopy outlined above have led not only to advancements in label-free Raman spectroscopy but also to the development of SERS, a highly sensitive advancement of conventional Raman spectroscopy. A variety of different metal surfaces can be used to provide surface-enhanced Raman but typically gold or silver surfaces which are plasmonic in nature are favored. This allows maximum electromagnetic enhancement due to the plasmonic coupling with the frequency of light used to excite the molecules being interrogated. Silver and gold can be used in a variety of different forms including electrodes, roughened metal films, vapor deposited films or very routinely as NPs. Silver or gold NPs are easily prepared through a metal reduction process which is scalable and controllable to give defined shapes and sizes of plasmonic NPs with degrees of enhancement that can be tuned to fit different laser excitations and also different surface chemistries. Most relevant to this review is the fact that they can be generated for use in vivo, functionalized with targeting molecules such as antibodies and tuned to work across a wide range of wavelengths from the visible to the near IR. SERS NPs can be used in a multiplexed manner by incorporation of spectrally distinct reporter molecules into the NP (Raman reporters). These are often dyes that are carefully selected so that when the obtained spectra from multiple Raman reporters (and hence different NP conjugates) are overlaid, each distinct reporter/ NP can be distinguished by a spectral peak (or peaks) unique to that particular reporter . This also allows quantification of the signal based on the intensity of the peaks obtained following comparison with a reference standard and subtraction of the background signal from the medium/tissue. These unique properties have given rise to a number of potentially promising biological applications of SERS and the current use and future potential of SERS for imaging inflammation will be discussed in the following sections.

Clinical use
Raman spectroscopy/SERS are not yet clinically approved for the diagnosis or treatment of disease. To our knowledge, the only current clinical uses within the biological sciences is for real-time bacterial strain identification using commercially available Raman microscopes (Eberhardt, Stiebing, Matthaus, Schmitt, & Popp, 2015) and an in vitro multiplex SERS-PCR assay for fungi identification (White et al., 2014).

Pre-clinical use and in-human trials
The major use for vibrational spectroscopy as a clinical diagnostic tool in immune disorders can be envisaged either as a tool for screening biological fluids or biopsy samples for disease specific molecular signatures or as a tool for in vivo imaging via molecular probes e.g. catheters introduced into vessels to monitor atherosclerotic plaque status or for monitoring the condition of the colon. Only the latter use will be discussed in this review and while progress is being made in developing vibrational spectroscopy for imaging inflammatory diseases, vibrational techniques are still at the proof of concept stage of pre-clinical development.
An area where vibrational spectroscopy might yield great clinical value is through integration of a Raman probe onto a catheter for intravascular use or for colonic examination of colitis and cancer. In 2005, Motz and colleagues developed an intravascular Raman probe for use on human patients. The first use of this intravascular Raman probe in humans was performed in 2006 on patients undergoing carotid endarterectomy and femoral artery bypass surgery. Quantitative characterization of the plaque stage was performed and compared with histological analysis of excised segments. Sensitivity and specificity of Raman detection was 79% and 85% respectively (Motz et al., 2006). Advancing on this methodology was the creation of a multi-modal imaging catheter developed by Scepanovic et al., 2011. Here, the authors employed a similar approach to Motz and colleagues on patients undergoing carotid endarterectomy and femoral bypass surgery. The fiber optic catheter consisted of a combination of reflectance IR spectroscopy, fluorescence (to measure autofluorescence from plaque components) and Raman detection. Each of these modalities provides complimentary information on the biochemical composition of the lesion including detection of collagen, elastin, lipids, hemoglobin and calcium. This was the first demonstration of a triple multi-modal system acquiring spectra from a vessel wall in human patients. The authors then used the system on ex vivo artery sections and used the combined biochemical data to stratify the stage and complexity of the lesion with the data subsequently correlated with gold standard histology. The data obtained from the multi-modal probe showed a correlated sensitivity of 93% and specificity of 72% compared with histological classification (Scepanovic et al., 2011) which are very promising preliminary results.
SERS has had rather limited use in vivo with earlier studies demonstrating proof of principle that multiple spectrally distinct NPs can be detected in vivo (Qian et al., 2008) and also one study demonstrated simultaneous detection of three distinct NPs targeted to 3 tumor biomarkers following direct injection into a mouse tumor (Dinish, Balasundaram, Chang, & Olivo, 2014), an important stepping stone to the in vivo multiplexed use of SERS. Our laboratory has been working to developing SERS as a multiplexing in vivo imaging tool for detection of vascular inflammation. In our first series of experiments, we were able to demonstrate that Pegylated NPs conjugated to an anti-ICAM-1 antibody could be employed to detect ICAM-1 expression on aortic sinus tissue sections from apoE −/− mice with a signal to noise ratio greater than that produced by immunofluorescence (McQueenie et al., 2012). We also supplemented this data with in vivo experiments using an intradermal inflammatory challenge model (LPS) in the ear pinna. LPS upregulates ICAM-1 expression in the ear vasculature and 24 h following LPS administration, FITC conjugated anti-ICAM-1 antibodies or anti-ICAM-1 functionalized SERS NPs were intravenously injected and non-invasively imaged by either two-photon fluorescence microscopy or SERS 30 min later. Both techniques detected upregulated ICAM-1 expression on blood vessel walls relative to isotype but as per the ex vivo studies, SERS showed superior sensitivity of detection (McQueenie et al., 2012). Recently, Gold Nanorods were deployed as SERS probes for the detection of adhesion molecules on inflamed macrophages (Pissuwan & Hattori, 2016). More recently we have utilized a SERS based system for the multiplex detection of VCAM-1, ICAM-1 and P-selectin on activated human macrophages and endothelial cells, human arterial and venous samples and a novel in-house developed in vivo model (Human adipose engraftment mouse model ( HA NSG)) that allows us to maintain viable human vascularized adipose tissue implants in an immunocompromised mouse (NSG mouse) in which we can test our SERS functionalized NPs (targeted to human specific inflammatory molecules) against our 3 inflammatory targets on human vessels in vivo following intravenous injection . In this work, we successfully obtained SERS spectroscopy and microscopy data from cultured inflammed endothelial cells, human arterial segments and most importantly from non-invasive SERS spectral analysis of adhesion molecules present within mice containing the transplanted human tissue following inflammatory challenge (Fig. 5). We have therefore demonstrated the feasibility of using the SERS-NP platform for simultaneous targeted multiplex detection of adhesion molecules in vivo. Adhesion molecules have been utilized in previous molecular imaging studies using established techniques such as PET and MRI; however, these studies predominantly explored single target molecule detection (Broisat et al., 2012;Dimastromatteo et al., 2013;McAteer et al., 2008;Nahrendorf et al., 2009). Recent advances have shown that the simultaneous identification and discrimination of 24 parameters is possible in vitro using an SRS approach (Wei et al., 2017) which makes extending the multiplexing potential of SERS in vivo a distinct possibility in the very near future thus allowing a more detailed interrogation of the molecular events that differentiate diseased from healthy tissue.
Endoscopic probes utilizing SERS NPs targeted to known cancer molecules on the luminal surface of the colon have been developed in tandem with fiber-optic Raman endoscopes for SERS detection of cancerous and pre-cancerous lesions. The most significant advance in this field is the creation of a human compatible, flexible Raman fiber optic device that can be used in the GI tract. It is a non-contact device and when applied in the colon has a working distance up to 25 mm from the colon wall. The device is also highly sensitive, being able to detect targeted NPs only sparsely bound to colonic wall cells (~240 NPs per cell) (Garai et al., 2015). It has also been demonstrated in similar studies that superficially introduced NPs into the colon do not pass over into the systemic circulation, even after 3 days and indeed most are naturally excreted within 24 h (Thakor et al., 2011), thus eliminating any potential off-site exposure. Assessing studies that have utilized either fluorescent or SERS NPs for detection of biomarkers using endoscope techniques has demonstrated to date a detection sensitivity of b1 f. for SERS compared with ≥ 10 f. for fluorescence-based methods. Therefore, SERS offers not only superior sensitivity but also the ability to multiplex a greater number of markers using a single monochromatic laser (Laing et al., 2017).

Theranostics
As introduced in Section 2.4, photothermal therapy utilizing NP size agents can offer a viable method for inducing cell death. NP designs including single-walled carbon nanotubes (SWNTs) and gold nanorods are particularly effective in inducing cell death in vivo (Chakravarty et al., 2008;Dickerson et al., 2008;von Maltzahn et al., 2009). While these studies were oncology focused, SWNTs are also readily uptaken by macrophages in carotid arteries in mice following carotid ligation. Following exposure of excised arteries to NIR laser excitation, macrophages engulfing SWNTs undergo photothermal cell death (Kosuge et al., 2012). The induction of apoptosis only following targeted excitation with NIR light provides a high level of specificity in targeting only lesion resident macrophages. Further refinements of photothermal therapy may include the use of multi-modal NPs to guide photothermal therapy and monitor outcomes. Gao and colleagues developed a gold NP (nanostar configuration that is highly amenable to heat absorption and with high sensitivity of detection (1 fM) that was surrounded with Gd chelates making the particle amenable to both SERS and MRI and with no observed toxicity out to 30 days following administration in mice (Gao et al., 2015). Remarkably, photothermal NP therapy has already been trialed in human atherosclerotic patients in the Plasmonic Nanophotothermal Therapy of Atherosclerosis trial (NANOM-FIM; ClinicalTrials.Gov identifier: NCT01270139) performed between 2007 and 2012. The combination of intravascular or transcutaneous NIR laser excitation with macrophage targeted NPs resulted in a reduction in total plaque volume, no adverse events throughout the 1 year follow up and increased survival in the treatment group (Kharlamov et al., 2015).
Another therapeutic avenue where SERS could be implemented is for delivery of anti-inflammatory drugs to the site of pathology and subsequent monitoring of drug release. The use of NP-drug conjugates that Fig. 5. In vivo SERS-BFNP molecular imaging of adhesion molecules. Following engraftment of human adipose, HA NSG mice were allowed to recover for 3 weeks. Mice were then injected intravenously with 5 μg of human recombinant TNF-α 4 h prior to receiving an intravenous injection of BFNP. (A) Following SERS-BFNP molecular imaging, adipose grafts were excised and immunofluorescently stained for human (red) and murine (green) CD31. (B) Isotype control, ICAM-1, VCAM-1, and P-selectin staining are also shown in white counterstained with human CD31 (red). Nuclei were counterstained using Hoechst 33342 (blue). (C) To conduct SERS-BFNP molecular imaging, HA NSG mice were anaesthetized and their adipose grafts noninvasively analyzed in vivo using SERS spectroscopy. (D) SERS spectra were acquired from mice that received a mixture of isotype-PPY, -BPE and -PYOT (blue spectra), or anti-Pselectin-PPY, anti-ICAM-1-BPE, anti-VCAM-1-PYOT BFNP (red spectra). The spectra shown are from 5 isotype vs. 5 targeted mice, with each spectrum acquired from a different mouse. (E) In addition to immunofluorescence microscopy, excised adipose grafts were analyzed using SERS microscopy. Detection of BFNP from sections of adipose tissue isolated from HA NSG mice that received anti-ICAM-1 (purple), anti-VCAM-1 (red), and anti-P-selectin (blue) (upper panels) or Isotype-BPE (purple), Isotype-PYOT (red), and Isotype-PPY (blue) (lower panels) are shown superimposed on darkfield tissue images alongside a magnified image of Raman maps from the scanned areas (black boxes). The colored circles in the Raman map ((E) upper panel) correlate to the acquired spectra shown in (F) above their respective reference spectra. The optical image in (E) is a darkfield image. Scale bars: (A) = 1000 μm; (B) = 100 μm; (E) = 20 μm. emit contrasting optical signatures pre-and post-drug release that can be detected by SERS offers the potential for measuring drug release under acidic pH conditions, such as found in lysosomes where internalized drug-NP conjugates are trafficked to following endothelial cell adhesion molecule targeting (Kelly et al., 2005;Muro et al., 2003;Ricard, Payet, & Dupuis, 1998;Serrano, Bhowmick, Chadha, Garnacho, & Muro, 2012). SERS-functionalized gold NPs carrying pH triggered doxorubicin release have successfully been targeted to cancer cells and drug delivery was confirmed by SERS tracking (Song, Zhou, & Duan, 2012). A similar method employing doxorubicin conjugated to gold NPs via a pH sensitive hydrazine linker was employed to also target doxorubicin to cancer cells (Kang, Afifi, Austin, & El-Sayed, 2013). Such an approach has not yet been applied to vascular endothelial cells or to deliver anti-inflammatory drugs but the utility of SERS sensitive NPs in such a context creates a potent dual-modality tool for both delivering and tracking therapeutic compounds to inflammatory sites.

Advantages and barriers to clinical translation
Raman probes either alone or on multi-modal catheters can map the biochemistry of the plaque with good sensitivity and specificity. The system created by Motz et al., 2005 was the first real-time imaging system for measuring Raman spectra intravascularly. In those studies, the entire duration of spectral acquisition, data processing and diagnostic read out was less than 5 s, although the in vivo acquisition was less than 1 s. One limitation of the Raman probe used by Motz et al. (2006) and the multi-modal system used by Scepanovic et al. (2011) is that they don't provide a real-time image of the vessel wall in the way that IVUS and OCT do. Therefore, the goal would be to combine a fast high resolution real-time imaging technique such as OCT (which can pinpoint areas interest) with Raman or a multi-modal system which can acquire the detailed molecular information. This would yield a highly powerful tool for investigating intravascular or colonic pathologies.
SERS is still at an early stage of development in terms of application to in vivo imaging, yet it has a number of strengths that may address the limitations of current imaging modalities. These can be outlined as follows: 1) Fast acquisition times; 2) high sensitivity, (enhancement over Raman can be as much as 10 10 -10 11 and even higher when the laser frequency is in resonance with the SERS molecule (Guerrini & Graham, 2012;McNay, Eustace, Smith, Faulds, & Graham, 2011)); 3) Low toxicity of AuNPs in animal models and early human trials; 4) Not susceptible to photobleaching; 5) Deep tissue imaging possible (up to 5 cm reported by Asiala, Shand, Faulds, & Graham, 2017); 6) Compatible with a broad range of fluorophores in the NIR window; 7) Multiple Raman reporters can be excited at a single wavelength; 8) Narrow emission spectrum allow accurate deconvolution of the data and high multiplexing; 9) Portable systems now available expanding the use beyond specialized research labs; 10) NPs can be configured for theranostic use such as photothermal therapy or as a site specific drug release vehicle.
In terms of barriers to translation, the ability to accurately deconvolute spectral data is a key requirement for all forms of vibrational spectroscopy. As the complexity of the sample matrix increases, more sophisticated mathematical and statistical data analysis routines are required to provide clear and concise data from the spectroscopic signals. The quantification of Raman and SERS data is a rapidly progressing area of research with highly promising results indicating SERS spectra can be quantified accurately using multiple mathematical models including Bayesian modeling (Zhong, Girolami, Faulds, & Graham, 2011) which we have successfully demonstrated shows a linear increase in proportion to increasing NP concentration . When aiming for the use of SERS NPs in human patients, the toxicity of the NPs must be accurately known including their residence times and fate within the patient (Cicha et al., 2018). Studies will need to address where SERS NPs accumulate, how this relates to NP size, shape and functionalization and any possible degradation prior to excretion. Gold per se is not a toxic metal and has been used in medicines for many years, for example gold based compounds are used to treat RA (Alkilany & Murphy, 2010); however, the toxic status of nanoscale gold particles remains obscure with differential toxic effects found in cellular and animal models depending on size, shape, charge, shell material and dose of NPs administered (Fratoddi, Venditti, Cametti, & Russo, 2015;Gerber, Bundschuh, Klingelhofer, & Groneberg, 2013). Shell material such as PEG has shown to be biocompatible and displays very low toxicity in human patients (Alexander, Ajazuddin, Khan, Saraf, & Saraf, 2013), being used in many pharmaceutical formulations around the world to improve drug residence times. PEG also reduces the toxicity of the NP by encapsulating the gold thus bare gold is not freely available in the circulation as well as acting as an anchor for conjugating targeting antibodies or peptides. Despite mixed pre-clinical results to date regarding the toxicity of NPs, silica-gold pegylated NPs have progressed to human clinical trials for thermal ablation of cancer (Stern et al., 2016). A safety evaluation trial in 22 prostate cancer patients who were administered an intravenous infusion of 'Aurolase' NPs and monitored for 6 months revealed only 2 minor adverse events related to the therapy. Interestingly and importantly, the FDA classified those NPs as a medical device rather than a drug (Stern et al., 2016) which reduces the time to regulatory approval. Multi-modal SERS NPs are also being developed. In an excellent study by Dinish et al., 2015, the authors created targeted NPs containing a NIR dye detectable by both SERS and photoacoustics in an in vivo mouse tumor model at a depth of several cm and at μM resolution allowing precision location of the tumor. The authors noted an accumulation of NPs in the liver and spleen but were cleared within 72 h, which is encouraging in terms of a lack of organ retention (Dinish et al., 2015). More generally, the creation of multi-modal SERS probes combined with modalities that offer visualization of tissue will greatly facilitate the location and optimum time point to initiate SERS data acquisition. To conclude our discussion on SERS, it is clear that SERS has many intrinsic advantages that may meet clinical needs currently unserved by classical imaging modalities, yet it will require substantial investment of resources across multiple disciplines to determine if these inherent advantages can be translated to a clinically useful imaging modality. 10 −3 -10 −5 Body-wide Good spatial resolution, non-invasive, high soft tissue contrast, no radiation exposure, low toxicity tracers with some already clinically used Poor sensitivity, long imaging times often required, poor signal to noise ratio, lack of multiplexing outside of specialist CEST MRI. b10 12 b5-6 cms High sensitivity, no radiation exposure, superior penetration depths compared with other optical modalities, low cost, anatomic and molecular information in a single scan, high speed real-time imaging, can be used on catheters for invasive imaging, multiplexing with MSOT, potential to act as a drug delivery theranostic tool Poor penetration depths when compared with current clinical imaging modalities, tracers not approved for human use, data acquisition and processing still to be optimized, less multiplexing capacity than vibrational spectroscopy Pre-clinical (Early clinical trials) Raman spectroscopy b1 μM to several hundred μMs~1 0 −5 to 10 −6 for analytes in solution ⁎

Several mms
Label free analysis, low cost, portable, real-time or near real-time imaging, high spatial resolution, able to discriminate a wide set of molecular structures, can be used on catheters for invasive imaging Low sensitivity leading to long imaging times, poor imaging depths restricting use to ex vivo or catheter probes, requires complex chemometric analysis to separate analytes from spectra

Summary and conclusions
To summarize, an ideal molecular imaging modality (either alone or as a hybrid with another) should possess high sensitivity and spatial resolution, fast imaging times, be able to image structures at depth within the body (unless utilized solely as an invasive catheter), be quantifiable and reproducible, have low economic burden and have the ability to assess multiple targets in a single scan. For successful clinical translation, short imaging times, patient comfort, accuracy, reproducibility, invasiveness of the procedure and necessity for hospitalization should be carefully considered. Highly specialized imaging modalities create an additional barrier by requiring trained reporting staff, thus wider application is easier with more approachable methods, such as US.
In a world where medical authorities and institutions demand greater value for money, systems that can be brought to clinical specifications with less regulatory hurdles are going to be highly prized. Of the current clinical modalities, CEUS and MRI may compete as safe, costeffective molecular and theranostic imaging tools, with similar capabilities and limitations. CEUS with more extensive use of microbubble contrast agents and targeted microbubbles now in clinical trials is further on this development path than MRI, although as 19 F MRI progresses into human use, only then will it be possible to judge the respective merits of these two modalities as potential molecular imaging tools that bring added diagnostic value over current systems. There appears little enthusiasm to move PET to clinical targeted imaging of disease biomarkers beyond oncology, but this may not be a permanent restriction. However, PET/SPECT will always be constrained by radiation exposure and the lack of multiplexing. The potential for clinical molecular imaging of emerging technologies is even harder to judge with any certainty given the scarcity of in-human (and even pre-clinical when it comes to the most recent technologies such as MPI and SERS) testing and the lack of completed clinical trials. We would suspect PAI will accelerate quicker than the other pre-clinical modalities and there is much hope within the pre-clinical academic community that it can translate to a clinical molecular imaging device. The application of SERS in vivo is just beginning to be realized with only a handful of publications in this area hence the technology is very early stage. However, its powerful set of attributes will continue to be tested in animal models and human tissue samples to build the robust, standardized quantification models that are necessary to make datasets consistent and comparable thus allowing progression to more rigorous diagnostic testing trials. We believe that the next decade will be key for all forms of vibrational spectroscopy in demonstrating whether they have a place in the clinic.
Major barriers to clinical translation of a molecular imaging modality include safety, cost, overcoming technological hurdles in signal acquisition and processing and performance versus established technologies/ clinical utility (Table 5). As demands on healthcare services increase, then the cost of implementing new technologies has to have a high threshold of clinical value. Although development costs are cheaper for a diagnostic imaging agent, costs are still substantial due to the lengthy development process because regulatory authorities treat molecular imaging agents as they would any other drug. An advantage of PET tracer development in the US -due to their trace amount administered dose -is that an exploratory investigative new drug application (eIND) pathway is available which reduces the time between preclinical and first in-human trials due to less stringent requirements for pre-clinical animal testing (Abou-Elkacem et al., 2015). This pathway is not open to other novel contrast agents such as targeted microbubbles or metallic NPs (Cicha et al., 2018). However, it is quite conceivable that as regulatory authorities become more convinced of the general safety of a contrast agent then expedited routes to clinical testing may be available if only the targeting ligand is changed on an otherwise previously tested safe contrast agent. Validation of this is emerging in CEUS with the KDR targeted microbubble used in a clinical trial in breast cancer showing excellent patient tolerability, comparable with a similar nontargeted microbubble. It is also interesting to note that the FDA classified the first pegylated gold NP therapeutic agent (Aurolase) as a medical device rather than a drug, reducing the burden and time towards regulatory approval. Whether this will remain the case for targeted gold NPs, such as those utilized in SERS or PAI is unclear as regulatory approval has not been granted for these imaging agents. At present, photoacoustic studies in humans rely on endogenous contrast agents or approved fluorescent molecules such as ICG (Valluru et al., 2016). Regardless, the lack of observed adverse events in patients treated with Aurolase could be a key milestone in the subsequent translation of metallic NP based agents into further human trials based around diagnostic imaging.
When considering costs, it is also important to keep in mind what the primary goal of molecular imaging is, namely to identify disease at an early stage or better predict its evolution and to monitor response to therapeutics. In this regard, clinicians require more robust integrated patient data that may include patient history, clinical scores, serum biomarkers and real-time, high resolution molecular imaging that can not only add clinical value, but better inform on the underlying cellular and molecular mechanisms in a patient-specific manner. The current inability to quantify individual disease progression in inflammatory conditions impairs our therapeutic choices and prevents personalized medicine. Our current approach to capture inflammatory processes is outdated, cumbersome and poses a substantial clinical limitation to precision medicine. Therefore, molecular imaging has the potential to bring many benefits to the clinic including improved patient outcomes, reduced diagnostic costs, reduced physical and emotional burden on patients and further cost savings to Health Authorities via the reduced demand for surgery and ineffective treatments due to improved decision making arising from the results of molecular imaging.

Funding sources
Our lab is also supported by the British Heart Foundation grants [