Molecular imaging in TB: from the bench to the clinic

Despite all efforts, tuberculosis (TB) still constitutes a serious global health threat with 9.4 million new cases and 1.7 million deaths worldwide in 2009 (World Health Organisation,2010). Furthermore, an estimated one third of the worlds’ population is infected with the bacterium responsible, Mycobacterium tuberculosis. The main handicaps in fighting TB include a vaccine which works poorly in the most affected populations, and an arduous treatment regimen, involving a combination of several drugs taken over many months. This is further complicated by the emergence of multi-drug resistant (MDR) and extensively drug-resistant (XDR) M. tuberculosis strains, which require even longer treatment times with less well-tolerated drugs. Eradication of TB will require the development of new drugs and vaccines, alongside improved methods for diagnosis and monitoring treatment efficacy. With the vast burden of disease falling in resource poor settings, the challenge will also be to develop methodologies that can be deployed with minimal investment in infrastructure, maintenance and staff expertise. Recent decades have seen the emergence of the new discipline of molecular imaging. In essence, molecular imaging enables the non-invasive visualisation, characterisation, and quantification of biological processes taking place within intact living subjects, be it a mouse or man (Filippi & Rocca, 2011; Horky & Treves, 2011; Pysz et al., 2010; Sandhu et al., 2010). Imaging has long been applied to managing TB; simple chest x-rays have allowed clinicians to visualise TB in people for over a century (Singh & Nath, 1994). However, the new molecular imaging techniques are revolutionising medical research, with the potential to translate into significant changes in clinical practice. In this chapter we describe the new generation of imaging modalities and how these are being applied to eradicating TB, from the laboratory bench and in to the clinic.


Introduction
Despite all efforts, tuberculosis (TB) still constitutes a serious global health threat with 9.4 million new cases and 1.7 million deaths worldwide in 2009(World Health Organisation, 2010. Furthermore, an estimated one third of the worlds' population is infected with the bacterium responsible, Mycobacterium tuberculosis. The main handicaps in fighting TB include a vaccine which works poorly in the most affected populations, and an arduous treatment regimen, involving a combination of several drugs taken over many months. This is further complicated by the emergence of multi-drug resistant (MDR) and extensively drug-resistant (XDR) M. tuberculosis strains, which require even longer treatment times with less well-tolerated drugs. Eradication of TB will require the development of new drugs and vaccines, alongside improved methods for diagnosis and monitoring treatment efficacy. With the vast burden of disease falling in resource poor settings, the challenge will also be to develop methodologies that can be deployed with minimal investment in infrastructure, maintenance and staff expertise.
Recent decades have seen the emergence of the new discipline of molecular imaging. In essence, molecular imaging enables the non-invasive visualisation, characterisation, and quantification of biological processes taking place within intact living subjects, be it a mouse or man (Filippi & Rocca, 2011;Horky & Treves, 2011;Pysz et al., 2010;Sandhu et al., 2010). Imaging has long been applied to managing TB; simple chest x-rays have allowed clinicians to visualise TB in people for over a century (Singh & Nath, 1994). However, the new molecular imaging techniques are revolutionising medical research, with the potential to translate into significant changes in clinical practice. In this chapter we describe the new generation of imaging modalities and how these are being applied to eradicating TB, from the laboratory bench and in to the clinic. processes longitudinally within the same subjects, a powerful tool indeed for elucidating host-pathogen interactions and treatment efficacy. A number of imaging modalities have emerged, which vary in their methods of image generation, spatial resolution, depth penetration and detection thresholds (Table 1). As a result each modality has different advantages and disadvantages (Table 2), suggesting the techniques should be used to complement each other to answer specific research questions.

Modality
Image generation Spatial resolution Depth penetration

Computed Tomography (CT)
CT imaging combines low-dose x-rays and computing to produce reconstructions of the internal organs and tissues. This is possible because diverse tissue types differentially absorb x-rays as they pass through the body. CT is not a molecular imaging tool per se, but can provide important information on anatomical changes which arise as a result of disease processes. Widely used in clinical settings, there are now a number of miniaturised machines suitable for scanning of small animals (often referred to as micro-CT). To collect data, the subject is placed on a motorised table, which then moves into the lead-encased CT machine. Inside, an x-ray source and a set of x-ray detectors rotate 360 degrees around the subject in synchrony. At every angle, the detectors record the x-rays passing through the subject to provide a digital projection which is collected and sent to a computer. The x-ray source produces a narrow, fan-shaped beam, with widths ranging from 1 to 20 mm. In axial CT, which is commonly used for head scans, the table is stationary during a rotation, after which it is moved along for the next slice. In helical CT, which is commonly used for body scans, the table moves continuously as the x-ray source and detectors rotate, producing a spiral or helical scan. Clinical machines typically have multiple rows of detectors operating side by si de , so that many slic es (cu rrently u p to 64) can be imag ed simu ltaneou sly , reducing the overall scanning time. As an alternative to the fan-shaped x-ray beam, small animal scanners may instead use a cone-shaped beam, where the scanned subject is captured completely in one rotation, speeding up the imaging process. The data are processed by computer to produce a series of image slices representing two-dimensional (2D) or three-dimensional (3D) views of the target organ or body region.  Massoud & Gambhir, 2003) The spatial resolution of CT is primarily limited by scanning times, the size of the x-ray source, and the sensitivity of the detection system. In addition, CT has relatively poor soft tissue contrast; generally, iodinated molecules are applied as contrast agents, owing to the high x-ray absorption coefficient of iodine (McClennan, 1994). Current iodine-based contrast agents have several limitations, including adverse reactions, renal toxicity, vascular permeation and rapid renal clearance resulting in limited imaging times. As a result, alternative contrast agents have been suggested, such as polymer-coated Bi 2 S 3 (Rabin et al., 2006) or gold nanoparticles (Hainfield et al., 2006). Indeed, gadolinium chelate-coated gold nanoparticles have been reported as dual imaging probes for CT and magnetic resonance imaging (MRI) (Alric et al, 2008). Tissue contrast can also be improved by using a dualenergy x-ray method in which the projection data are acquired using two different x-ray spectra (Taschereau et al., 2010). However, one of the major limitations of CT is radiation exposure, and while the doses are low, they are not negligible and this can limit repeated imaging of the subject.

Magnetic Resonance Imaging (MRI)
MRI is based on the interactions of atoms and molecules in a tissue of interest, upon exposure to a magnetic field. In addition to providing detailed structural images, MRI can obtain physiological information through the use of specific contrast agents. While the proton 1 H is most widely used in MRI, due to the abundance of water within soft tissues, other paramagnetic atoms such as 13 C, 17 O, 19 F, 23 Na and 31 P are also useful. Within an MRI scanner, a strong 'coiled' magnet produces a magnetic field with a gradient in the X, Y and Z directions, which causes nuclei to align themselves. The device also contains a radiofrequency (RF) coil which is used to produce a temporary RF pulse, resulting in a change in nuclei alignment. Following the pulse, the protons return to their baseline orientation (known as relaxation) which is detected as a change in electromagnetic flux.
The behaviour of the energy inserted into the system is described by two relaxation constants: the longitudinal relaxation time (T1) or the transverse relaxation time (T2). Different tissues have different relaxation times and this can be used to produce endogenous contrast between different tissues. Addition of exogenous contrast agents can further enhance tissue contrast by selectively shortening either T1 or T2 in a tissue of interest. According to their magnetic properties, contrast agents can be classified as paramagnetic (for example, gadolinium based agents) or superparamagnetic (for example, iron oxide nanoparticles) (reviewed in Geraldes & Laurent, 2009). Depending on their biodistribution patterns, different contrast agents can also be utilised to image specific anatomical regions. MRI is becoming widely used in both clinical and preclinical settings, with dedicated MRI machines available for humans and rodents. An advantage of MRI is that it does not involve ionising radiation, has unlimited depth penetration and good soft tissue contrast. However, it is expensive and scanning times are typically long, from minutes to hours.

Positron Emission Tomography (PET)
PET imaging involves the visualisation of a radiotracer, a biomarker labelled with a positron emitter. The positron emitters typically used are isotopes with short half-lives (several hours to a few minutes), such as 11 C, 13 N, 15 On and 18 F. Radiotracers are typically made to reflect compounds normally used by the body, such as glucose or ammonia, or molecules that bind to specific receptors. Once the radiotracer is injected into a subject, it therefore distributes based on its similarity to the original biomarker compound. The most commonly used radiotracer is an analogue of glucose labelled with 18 F, [ 18 F]-2-fluoro-deoxy-D-glucose, ([ 18 F]-FDG). A major advantage of PET imaging is that it can be used to trace the fate of any compound, provided it can be radiolabeled with a PET isotope. As a result, the processes that can be probed using PET imaging are virtually limitless, and radiotracers for new target molecules and processes continue to be developed. For a summary of available PET radiotracers see Pysz et al., 2010. Dedicated clinical and small animal PET scanners are now available.
www.intechopen.com PET imaging is based on the fact that the incorporated radionucleotide undergoes positive decay and emits a positron. The positron travels a few mm before it annihilates with an electron to emit a pair of photons moving in approximately opposite directions. These photons are then detected by the scanning device. As the photons are travelling at approximately 180º to each other, it is possible to localise their source along a straight line of coincidence known as a line of response (LOR). The distribution pattern of the LORs is then used to reconstruct an image of the radioactivity distribution within the subject. One minor limitation of utilising photons is that they are differentially attenuated as they traverse different thicknesses of tissue. This attenuation results in the reconstruction of structures deep within the body as having falsely low uptake of the radiotracer. However this attenuation can be corrected for by combining PET with CT imaging.
Despite the great promise of PET imaging, there are a number of significant disadvantages. One is the use of ionising radiation, although this is minimised by the use of radioisotopes with short half-lives. However, these short half-lives require both costly cyclotron generators and chemical synthesis apparatus within close proximity to the scanning facility for the production of the radiotracers. This certainly limits the use of the technology within resource poor settings. Furthermore, scanning times are typically long, from minutes to hours, and the technique provides low spatial resolution.

Single Photon Emission Computed Tomography (SPECT)
Like PET, SPECT imaging is based on the distribution and uptake of a radiolabelled tracer after injection into a subject. Unlike PET tracers, SPECT radionucleotides undergo radioactive decay and emit -rays of a particular energy, which are then captured by an external detector. A number of 2D projections are captured from multiple angles which, when combined, form a 3D image. Radiotracers based upon radioactive metals, such as 111 In, 188 Re, 131 I, and 133 Xe, are often used. For a summary of available SPECT radiotracers see Pysz et al., 2010. PET and SPECT imaging share many of the same advantages and disadvantages. However, while PET is more sensitive, SPECT imaging is much cheaper largely thanks to the availability of different radiotracers which are longer lived and easier to obtain. Moreover, different SPECT radiotracers have different energies enabling multiple tracers to be used to image different processes.

Optical Imaging
The electromagnetic radiation we refer to as light undergoes a range of interactions when propagating through tissue. Importantly, these interactions depend on the structural arrangement and physical properties of the micro-environment. Such interactions have led to the development of the field of optical imaging which encompasses a wide variety of methods and approaches (Table 3), from visualising tissue anatomy on the microscopic scale (Zonios et al., 2001) to the 3D localisation of a photonic signal in whole animals using fluorescence molecular tomography (Ntziachristos, 2006).
In this chapter we will focus on biophotonic imaging (BPI), a preclinical imaging technique based on the ability of light to travel through flesh. This principle is easily demonstrated by placing a torch underneath ones hand and observing the light emerging through the fingers. BPI involves the detection of visible light which arises from either the excitation of a  Bioluminescence arises from the oxidation of a substrate (a luciferin) by an enzyme (a luciferase), which usually requires energy (in the form of FMNH 2 and ATP) and oxygen. Luciferin and luciferase are generic terms as none of the major classes share sequence homology. Most widely studied are the systems belonging to luminous beetles in the family Lampyridae (such as the firefly Photinus pyralis), the sea pansy Renilla reniformans, the marine copepod Gaussia princeps and numerous luminous bacteria (such as Vibrio sp. and Photorhabdus luminescens). In contrast, fluorescence arises when a fluorescent compound is irradiated with light of a suitable wavelength. This leads to the transition of an electron in the molecule to a higher energy state, a process known as excitation. This process is almost instantaneous, taking around 10 -15 seconds. Upon return of the electron to a lower energy level (around 10 ns), light of lower energy is emitted, giving the fluorescent signal.
Although the emitted light may be dim, it can be detected externally using sensitive photon detectors such as those based on cooled, or intensified, charge coupled device (CCD) cameras, mounted within light-tight specimen chambers. As light passes through a range of tissue types (including skin, muscle and bone) it is possible to observe and quantify the spatial and temporal distribution of light production from within living subjects (N. Andreu et al, 2011). In general, imaging of luminescence is much more sensitive than imaging fluorescence as a result of better signal-to-noise ratios. This is mainly due to the high levels of background fluorescence in vivo compared to luminescence, due to endogenously produced fluorophores such as keratin, porphyrins, NAD(P)H, collagen and elastin (Troy et al., 2004). A major limitation of BPI is the limited depth penetration through tissue. Hence BPI is currently only applied to imaging small animals, although visualisation of bioluminescence from within infant monkeys (the long-tailed macaque, Macaca fascicularis) has been reported (Tarantal et al., 2006). Alternatively, the light could potentially be detected internally using an endoscopic device, such as reported by Hsiung and colleagues www.intechopen.com to image colonic pathology (Hsiung et al., 2008). The advantages of BPI are that it is inexpensive, sensitive and requires short imaging times.

Use of molecular imaging in animal models of TB
M. tuberculosis, the infectious agent of TB, can infect many animals in addition to its natural human host. Although the study of TB in patients is extremely useful, a detailed analysis of the pathogenesis and the interactions of M. tuberculosis with the host requires the use of well-defined models that can be infected in a controlled manner. Furthermore, animal models can be easily manipulated, can be used in statistically significant numbers, and the results are obtained in a relatively short time frame. In addition, they are particularly useful in drug and vaccine efficacy testing before moving the most promising candidates to clinical studies.
For both practical and economical reasons, laboratory mice remain the most extensively used animal model of TB: they are easy to manipulate and house, there is a wide range of mutant and genetically modified strains, and there are many immunological reagents available. However, latent infection is difficult to achieve in the mouse model, and the pathology, with poorly organised granulomas, differs considerably to that observed in humans. By contrast, guinea pigs and particularly rabbits display a spectrum of pathology that better represents the human disease. Moreover, guinea pigs are extremely susceptible to M. tuberculosis infection and relatively inexpensive compared to other larger animal models, which makes this model very useful for vaccine efficacy studies. Even so, studies with guinea pigs and rabbits are limited by the narrow range of immunological reagents available. This is not the case in non-human primates, which are the closest model to humans in terms of pathology and disease development and therefore constitute the most relevant model to predict treatment and vaccine efficacy. Nevertheless, work with nonhuman primates presents many limitations regarding space requirements, animal availability, and costs. In summary, each animal model presents both advantages and disadvantages which must be carefully considered when designing a new study. A more detailed description of these animal models of TB can be found elsewhere (Dharmadhikari & Nardell, 2008;Flynn, 2006;Gupta & Katoch, 2005).
The use of animals in research is accompanied by ethical responsibilities and most countries promote the three Rs: replacement, reduction and refinement. Replacement refers to methods that avoid the use of animals, for example, in silico computer modelling, or using established human and animal cell lines and non-mammalian models such as the nematode Caenorhabditis elegans or the embryo of the zebrafish, Danio rerio. Reduction refers to methods which minimise the use of animals and enable researchers to obtain comparable levels of information from fewer animals or to obtain more information from the same number of animals, thereby reducing the future use of animals. Refinement refers to improvements to scientific procedures and husbandry which minimise actual or potential pain, suffering, distress or lasting harm and/or improve animal welfare. Molecular imaging is a very powerful tool for implementation of two of the 3Rs, refinement and reduction. Using traditional disease models, infected animals are sacrificed at defined time points and tissues excised for determination of pathogen numbers and localisation. In contrast, the nondestructive nature of molecular imaging allows the course of an infection to be monitored simply by repeated imaging of the same group of animals. Importantly, this allows disease progression to be followed with extreme accuracy, while allowing each animal to act as its own control. Furthermore, we have demonstrated that BPI can provide real time information on the effectiveness of the inoculation method (Wiles et al., 2007). As a result, errors in administration can be detected immediately (N. Andreu et al., 2011) and animals eliminated from further study -thus minimising any potential pain, suffering and distress for the animal and reducing variation by removing flawed scientific data.
One major drawback to working with M. tuberculosis is the slow growth of the organism. This lengthens the time required to carry out in vivo experiments extraordinarily, and delays the quantification of bacterial burdens by about four weeks, which is the time required for M. tuberculosis to form visible colonies on agar. Therefore, the use of molecular imaging to track infection dynamics in real time would be a major advantage as it would enable researchers to make on-the-spot decisions, shortening the length of the experiment if clear differences (for example, between control and vaccinated groups) were observed. There is, therefore, an increasing interest in the development of molecular imaging techniques in animal models of TB. Moreover, the developments and knowledge acquired through the use of these techniques in animal models may eventually translate into the clinic.

Computed Tomography (CT)
CT imaging has mostly been used as a complementary technique to PET and SPECT imaging (see sections 3.3 and 3.4), as it gives high-resolution anatomical information for a better localisation of the radionuclide signal. However, CT has also been evaluated as an imaging method on its own to assess disease burden in macaques (Lewinsohn et al., 2006). To this end, four animals were infected by bronchoscopy instillation of M. tuberculosis, and disease progression was monitored every four weeks clinically (weight, body temperature, complete blood count and erythrocyte sedimentation rate), immunologically (ELISPOT), bacteriologically (quantitative M. tuberculosis culture from bronchoalveolar lavage), and by CT imaging. In addition, a necropsy was performed at the end of the experiment (12 weeks post-infection) which included histopathology and bacterial burden quantification from selected organs. Clinical indicators failed to provide information about disease progression, as most of them were fairly constant through the whole experiment. Most bacterial cultures from bronchoalveolar lavage were positive, although some cultures were negative even though CT imaging and post-mortem analysis showed infection. Even bacterial cultures from post-mortem lung samples were not consistently positive, which was attributed to a non-uniform infection of the lungs and therefore biased tissue sampling. In contrast, CT imaging provided a reliable readout of disease progression in the whole lung and also allowed monitoring of other organs, such as the liver and spleen. Moreover, different types of lesions were observed, and progression of the lesions from small nodules to cavitation and necrosis was evident. CT findings were corroborated by post-mortem histopathology and, together with immunological monitoring, provided a non-invasive, accurate, and rapid assessment of TB in this animal model. It is important to note that even though a CT scanner was not available in the animal biosafety level 3 (BSL3) containment facility, the authors were able to image infected animals in a scanner localised within a non-containment facility, by transporting and imaging the anesthetised macaques in a box fitted with HEPA filters. This is a solution that has also been adopted for other imaging techniques like PET/CT, SPECT/CT and BPI (see below).

Magnetic Resonance Imaging (MRI)
To our knowledge, MRI was the first molecular imaging method reported for an animal model of TB, when Kraft and colleagues used the technique to assess lesion distribution and lesion numbers as an indication of disease burden in BCG-vaccinated and unvaccinated guinea pigs infected with M. tuberculosis by the aerosol route (Kraft et al., 2004). 3D lung images were reconstructed from images taken of 2 mm slices of formalin-fixed and agaroseembedded lungs, and lung volumes, lymph node volumes and total nodular burden were quantified. Small nodules were observed 15 days post-infection, which developed into granulomatous lesions 20 days later. Lesions were uniformly distributed in the lungs, which suggested that aerosol delivery of M. tuberculosis results in a homogenous infection. Additionally, lesions numbers supported the hypothesis that a single bacillus establishes a single lesion. In terms of vaccine efficacy, the authors found the same number of lesions in vaccinated and unvaccinated animals but the lesions were smaller in the vaccinated group, thus suggesting that BCG has an effect on disease development rather than on the initial establishment of the infection. All in all, they found that MRI was a useful method to assess disease burden in terms of lesion distribution, size and number. The main limitation was a low sensitivity when dealing with very small (< 1 mm) lesions.
More recently, the same laboratory used MRI to assess treatment efficacy in guinea pigs infected with M. tuberculosis (Ordway et al., 2010). The treatment had a dramatic effect on bacterial load with a 4-6 log decrease in viable counts (as determined by colony forming units [CFUs]) both in the lungs and lymph nodes in just 25 days. However, the effect on lesion burden, as quantified by MRI, was slower and could only be detected in the lungs after 50 days of therapy ( Figure 1). In addition, the lesions in the lymph nodes of the treated group were smaller, although the differences with the untreated group were only obvious at later time points. These results were corroborated by histological analysis, although the number of lesions in the lungs of the treated animals was already lower than in the control group by day 25.
MRI has also been applied to studies in non-human primates (Sharpe et al., 2009(Sharpe et al., , 2010. Disease burden in this animal model has been traditionally assessed by a range of ante-and post-mortem methods such as clinical signs (behaviour, weight, and body temperature), laboratory markers (haemoglobin levels, erythrocyte sedimentation rate, and immunology), chest x-ray, gross pathology, and histology. However, most of these methods are qualitative and subjective. Moreover, the most common alternative method, the quantitative estimation of total lesion numbers in the lung by manual counting, is laborious and particularly difficult in animals with more severe disease, as individual lesions become difficult to distinguish. Sharpe and colleagues used MRI and stereology (a statistical method that extracts quantitative information of a 3D structure from measurements made on planar sections of the material) to quantify lesion volume relative to lung volume in macaques infected with a range of doses of M. tuberculosis by the aerosol route (Sharpe et al., 2009). Similarly to what was previously seen in guinea pigs, the authors observed a uniform distribution of lesions in the lungs. In addition, the lesion-to-lung volume ratio increased with the infectious dose, and this ratio revealed subtle differences in the level of pulmonary disease and correlated well with other measures of disease burden. By contrast, methods such as gross pathology and chest x-ray were less sensitive and did not differentiate between the levels of disease in the animals exposed to the highest infectious doses. In conclusion, MRI together with stereology makes up a sensitive, quantitative, systematic and consistent method to assess disease burden in the macaque model of tuberculosis. Moreover, when MRI was compared with more traditional methods to measure vaccine efficacy, it was found that MRI combined with stereology was the only readout that distinguished between the unvaccinated and the vaccinated groups, and it was even able to show differences between survivor and non-survivor animals within the vaccinated groups, thus highlighting the sensitivity of the method (Sharpe et al., 2010).
In summary, the use of MRI appears to be a reliable method to assess disease burden in the lungs of M. tuberculosis-infected animals. However, it should be noted that the studies described here were performed on fixed lungs where the bacteria had been inactivated, as the use of MRI under BSL3 containment was not available. Initially, the whole lung was fixed and used for imaging to reduce sample error. As a result the tissue could not be used for other procedures, such as determination of bacterial load. However, the results discussed above illustrate that aerosol delivery of M. tuberculosis results in an even distribution of the lesions in the lungs and, therefore, samples can be taken and used for other techniques without compromising its reliability. Similarly to what has been done to image live animals by CT scanning, MRI of live animals could be done by using a sealed box with filters to transport the animals to the MRI facility and contain them during imaging. When available, MRI of live animals will allow longitudinal monitoring of disease progression, and real-time observation of vaccine and drug efficacy. The ex vivo results discussed here, together with the excellent soft tissue contrast of MRI and the development of faster MRI devices that reduce the artefacts induced by respiratory motion, suggest that in vivo MRI may become a very useful technique for the study of TB in animal models.

Positron Emission Tomography (PET)
Another technique which is gaining popularity in TB research is PET combined with CT imaging (PET/CT). The PET radiotracer [ 18 F]-FDG is used to image inflammation at the infection site, as it accumulates in inflammatory cells such as neutrophils and activated macrophages. This technology has been used to image TB infection ( Figure 2) and to assess drug treatment efficacy in mice (Davis et al., 2009b). The authors infected two strains of mice, BALB/c (which develop diffused granulomas) and C3HeB/FeJ (which develop wellwww.intechopen.com defined necrotic granulomas), and evaluated three different treatments: (i) first line tuberculosis regimen (rifampin + pyrazinamide + isoniazid), (ii) a more bactericidal regimen (rifampin + pyrazinamide + moxifloxacin), and (iii) a bacteriostatic regimen (ethambutol). The animals were imaged and CFUs obtained at different time points during the 12 weeks of treatment. Furthermore, one group of BALB/c mice was followed for 22 weeks after completion of the bactericidal treatments to assess relapse of the infection. For the imaging, anesthetised mice were contained in a sealed device with holes for passage of gases fitted with 0.22 m filters. In both mouse models, CFU counts perfectly reflected the efficiency of the three treatments being evaluated, with a faster decrease in bacterial numbers when moxifloxacin was used instead of isoniazid, and stabilization of bacterial burden when mice were treated with ethambutol only. The use of PET/CT imaging allowed differentiation between the bacteriostatic and bactericidal treatments, as the [ 18 F]-FDG activity was higher in the mice treated with ethambutol. However, unexpectedly, [ 18 F]-FDG activity was higher in mice treated with moxifloxacin than in those treated with isoniazid during the first four weeks of treatment. The authors suggested that this could be due to the limited statistical power of the study since only three mice per group were used, or that it could be an inherent limitation of using [ 18 F]-FDG, whereby an increased killing of M. tuberculosis would cause an increased Tumour Necrosis Factor (TNF)-mediated inflammation and therefore increased [ 18 F]-FDG activity even though bacterial numbers were decreasing. Relapse was detected in both groups of mice by PET imaging and by CFU counts. In summary, PET/CT allowed the non-invasive monitoring of disease progression in real-time. Moreover, individual lesions could be observed in the C3HeB/FeJ mouse model; as treatment response has been suggested to be lesion-dependent, the possibility of monitoring individual lesions would be very useful. However, it is important to take into account that this method does not specifically image infection but only measures inflammation which, as illustrated by the treatment results of this work, does not always correlate with bacterial burden. Nevertheless, this method has some advantages over using, for example, CFU counts: it uses a reduced number of animals, and the same animals can be repeatedly imaged which allows a more easy detection of untimed events such as relapse.
PET/CT imaging is also being used in non-human primates, although the results have only been presented in meetings and no peer-reviewed article has been published to date. For example, PET/CT has been used to monitor disease progression and drug efficacy in macaques (Lin et al., 2009). Using CT imaging, lesions as small as 1 mm were detected in the lungs and lymph nodes of infected animals. Moreover, lesion progression could be followed over time.

Single Photon Emission Computed Tomography (SPECT)
SPECT/CT has also been used for imaging of TB infection in mice (Davis et al., 2009a). The authors used the radiotracer 1-(2′deoxy-2′-fluoro--D-arabinofuranosyl)-  (Figure 3). The presence of the lesions was subsequently corroborated by histopathology. However, the limit of detection was found to be 5x10 6 to 1x10 7 CFUs, a rather high bacterial burden for mice infected with M. tuberculosis. The authors suggested that the sensitivity of the method could be improved by increasing the expression of TK in the bacilli or by using more sensitive (and expensive) radionuclides such as 123 I or 124 I. Other limitations of the technique include: the limited blood supply at the centre of the granulomas could limit accessibility to imaging substrates; TK requires ATP, which could be restricted in latent bacteria; and the presence of non-specific signal in tissues such as liver, gall bladder, or stomach, that either metabolize or excrete FIAU or its iodinated derivatives.

Biophotonic Imaging (BPI)
Bioluminescence imaging is one of the most widely used imaging techniques in the study of infectious diseases (N. Andreu et al., 2011). Luciferases have been used in mycobacterial research for more than 20 years; the two most widely used are the firefly luciferase (FFluc) and the luciferase of the bacterium Vibrio harveyi (LuxAB). Both luciferases produce light in the presence of a combination of a substrate and a cofactor, namely D-luciferin and ATP (for FFluc) and n-decanal and FMNH 2 (for LuxAB). As the co-factors are only found in live cells, the production of light by the luciferases provides a sensitive indicator of cell viability. The bacterial luciferase system has a major advantage when compared with the FFluc: the genes for the synthesis of the substrate are known and can be co-expressed with the luxAB genes as a convenient gene set (luxCABDE) that renders the bacteria autoluminescent, that is, no external addition of substrate is needed for light production. Light-emitting mycobacteria have been used as an easier and faster approach than commonly used methods to assess bacterial numbers in vitro and in macrophages, for example, in drug screening assays . The first approach to use luminescent mycobacteria in animal models consisted of measuring luminescence ex vivo in organ homogenates (Hickey et al., 1996). This method generated results in a much quicker time frame than using CFU counts and has been applied to drug and vaccine efficacy testing (Hickey et al., 1996;Snewin et al., 1999).
More recently, a recombinant M. bovis BCG strain expressing the bacterial luciferase enzyme LuxAB has been used to monitor mycobacterial infection in vivo (Heuts et al., 2009). In this work, only the luciferase genes were expressed and, therefore, the n-decanal substrate had to be injected before imaging. Although n-decanal is very toxic, the authors were able to deliver it dissolved in a mixture of olive oil and ethanol by injection into the mouse peritoneum. To assess the usefulness of the system, immunodeficient RAG2 −/− /cR −/− mice were intravenously infected with the luminescent BCG strain, and bioluminescence imaging was performed at different time points for 11 weeks. A signal coming from the spleen was detected four weeks post-infection, when the bacterial load was around 5x10 7 CFUs. The signal increased over time and extended to the abdomen of the animal but no signal was observed in the lungs, even though CFU counting showed a bacterial burden in this organ of 10 7 CFUs at eight weeks post-infection. However, luminescence was detected in the excised lungs, suggesting that tissue attenuation was responsible for the failure to detect the signal in the whole animal. The same luminescent BCG strain was also used to assess drug efficacy and the host immune response. A reduction in light emission, which paralleled the reduction in bacterial numbers, was observed in treated mice compared to the untreated mice, as well as in inmunocompetent BALB/c and T-cell reconstituted RAG2 −/− /cR −/− mice compared to immunosupressed RAG2 −/− /cR −/− mice. Therefore, bioluminescence imaging allows monitoring of mycobacterial infection in mice. However, the system was not useful for imaging infection in the lungs, and a toxic substrate had to be administered to the mice before imaging.
To overcome these difficulties, our group has recently optimised the expression of FFluc and the complete bacterial luciferase system in M. tuberculosis (N. Andreu et al., 2010). The resulting mycobacterial strains express either the optimised gene encoding FFluc (which is the brightest luciferase and uses a non-toxic substrate) or the optimised luxCABDE gene set from Photorhabdus luminescens (which results in autoluminescent strains that do not need the exogenous addition of substrate to produce light). Both M. tuberculosis strains were imaged in vivo in the lungs and spleens of infected mice (Figure 4), with limits of detection of around 10 5-10 6 CFU per lung and 10 5 CFU per spleen, whereas as few as 10 4 CFU can be imaged in the dissected organs (our unpublished results). Further work will assess the usefulness of these luminescent mycobacteria in drug efficacy testing and in other small animal models such as guinea pigs. Fluorescence imaging has had a more limited use in the study of infectious diseases, although it has been widely applied to other research fields such as cancer research (N. Andreu et al., 2011). In a first attempt to develop fluorescence imaging of M. tuberculosis infection, a GFP-expressing M. tuberculosis strain was used to infect mice and guinea pigs, and five weeks post-infection the lungs were imaged using a photon imager (Sugawara et al., 2006). Granulomas as small as 1 mm of diameter were detected, and the results were corroborated by histopathology examination. The same fluorescence technique was used to visualize granulomas in a latent model of TB in guinea pigs (Sugawara et al., 2009). In this case, the animals were subcutaneously infected with M. tuberculosis and the infection was followed for 10 months. No clinical signs of infection were evident in any of the animals for the length of the experiment, although the bacteriological analysis of lungs and spleens 180 and 300 days post-infection showed the presence of a few bacteria. Similarly, even though no granulomas were detected by gross pathology examination, microgranulomas were observed in the histological analysis. According to the authors, these small lesions corresponded to the fluorescent spots detected by photon imaging of sliced lungs and spleens. More work is still needed to validate these results and to be able to use this technology in vivo.
FPs that emit light in the far-red region of the spectrum are more appropriate for in vivo imaging than for example GFP, as red light is less affected by absorption and scattering www.intechopen.com when travelling through tissues. This is exemplified by the finding that as few as 10 5 CFUs of a BCG strain that expressed tdTomato (excitation 554 nm, emission 581 nm) can be detected after being subcutaneously injected into mice, in comparison to 10 7 CFUs of BCG expressing enhanced GFP (EGFP) (excitation 484 nm, emission 510 nm) (Kong et al., 2009). Therefore, far-red reporters show a lot of promise for fluorescence in vivo imaging of M. tuberculosis infection in animal models, and the expression of several red FPs in mycobacteria has recently been optimised (Carroll et al., 2010).
The strategies presented above use recombinant bacteria that express an exogenous FP. A much more versatile strategy consists of using an activatable fluorescent agent that is nonfluorescent in its native (quenched) state but produces fluorescence through enzymemediated release of its fluorochrome. This strategy has been widely used in cancer and inflammation research but, until recently, not in infectious diseases. Using a near-infrared fluorogenic substrate for -lactamase, an enzyme that is endogenously expressed by M. tuberculosis but not by eukaryotic cells, it is possible to detect 10 6 CFUs of M. tuberculosis subcutaneously injected in mice (Kong et al., 2010). The maximal signal was produced 48 h after substrate injection, and no signal was detected 48 h later, which suggests that repetitive imaging of the same animals can be done every 96 h. Surprisingly, the limit of the detection in the lungs of live animals was 10 4 CFUs, which is far lower than the limit of detection subcutaneously, even though the lungs are localised much deeper in the body. The signal was localised laterally, close to the armpit of the animal, and 3D fluorescence molecular tomography (FMT) and imaging of the excised lungs proved that the lungs were the source of the signal. The amount of fluorescence correlated with bacterial numbers when the imaging was performed 24 h post-substrate administration, but at later time-points substrate accumulation lead to a similar level of fluorescence independently of bacterial numbers. In addition, the technique was used to assess drug efficacy by imaging treated and untreated mice, showing that the signal increased in the untreated group while it decreased in the treated group. Although much work needs to be done to assess the usefulness of the technique, one can imagine many potential applications not only in in vivo imaging but also in vitro using fluorescence microscropy or FACS, as well as for TB diagnosis (e.g. detection of bacilli in sputum or imaging tuberculosis in patients). However, a limitation that needs to be considered is the fact that other bacteria, such as Pseudomonas aeruginosa or Staphylococcus aureus, also express -lactamase and therefore may give a false signal. Thus, alternative fluorogenic substrates that are activated by other endogenous enzymes are currently under study; for example, certain trehalose analogues that are substrates for the M. tuberculosis mycolyltransesterases Ag85A, Ag85B and Ag85C (Backus et al., 2011).

Radiography
The plain radiograph was first described in 1895 by Röntgen, at the same time that the TB pandemic was peaking in wealthy Western countries such as Victorian England. Consequently, there is a long experience and literature of plain radiographic imaging of TB, which has been reviewed previously (McAdams et al., 1995;J. Andreu et al., 2004;Curvo-Semedo et al., 2005). In summary, TB can cause a wide array of chest x-ray appearances, but classically causes consolidation and cavitation in the apices of the upper lobes. TB can also cause disease at the apices of the lower lobes, which appears in the mid-zone on chest radiograph ( Figure 5). However, TB can result in a wide range of other features, such as miliary disease with small millet seed-sized nodules throughout the lungs, pleural effusions, mediastinal lymphadenopathy and extrapulmonary disease. In the era of HIV infection, where the host immune system is compromised, the appearances of pulmonary TB are often atypical (Kwan & Ernst, 2011), ranging from classical cavitation to areas of pneumonia to a normal chest x-ray even in the presence of a high mycobacterial load. This illustrates the importance of the adaptive immune response in driving lung inflammation, resulting in consolidation and tissue destruction. As TB is treated, areas of consolidation tend to gradually resolve, leaving an area of fibrosis or scar tissue which persists for life. Cavities remain even after cure, because the lung cannot reconstruct the intricate extracellular matrix after it has been destroyed. However, it is well recognised that radiographic appearances may often worsen before they improve (Leung, 1999), and similarly some lesions may increase in size and density while others appear to resolve. Even in HIV negative patients with drug-sensitive disease, such "paradoxical" reactions may occur (Cheng, 2002). This demonstrates the different behaviour of inflammatory lesions even in the same patient, and one challenge for modern imaging techniques is to define the molecular mechanisms underlying this immune response to improve our understanding of what constitutes an effective as opposed to deleterious immune response to TB.

Computed Tomography (CT)
CT involves cross-sectional imaging of patients and so permits a much greater degree of resolution of anatomical structures, although it results in a higher radiation dose and higher cost than plain radiography. CT can demonstrate cavity formation with much greater sensitivity and will demonstrate subtle changes which may be missed on plain chest x-rays ( Figure 6 [left panel]). For example, filling of small airways with inflammatory debris may result in a "tree-in-bud" pattern ( Figure 6 [right panel], arrow), which should immediately alert the physician to the possibility of mycobacterial infection. Therefore, CT scanning provides information of changes at a much more precise anatomical level than plain www.intechopen.com radiography, but does not give information about the molecular events occurring at the site of infection. Fig. 6. CT imaging reveals a small left mid zone cavity (indicated by arrow) that was not visible on plain chest x-ray (left panel) and a tree-in-bud bronchial filling in a 33 year old man with pulmonary TB (right panel), which appears as branched opacities in the lung field adjacent to arrowhead.

Magnetic Resonance Imaging (MRI)
MRI provides the best imaging of the meninges and spinal cord, and so is useful in the diagnosis of cerebral TB, TB meningitis and paraspinal TB abscesses. MRI is a rapidly developing area, and so may emerge as a modality which can provide insight into molecular events in TB. The advantage of MRI is that it involves no ionizing radiation and provides excellent anatomical resolution. New MRI modalities to investigate inflammatory diseases are under development (Pirko et al., 2004), but these have not yet entered the clinical arena for investigation in TB.

Positron Emission Tomography (PET)
PET imaging is a widely used nuclear medicine technique which has the potential to study pathological events at a molecular level before extensive anatomical changes are observed on plain radiography. PET imaging is commonly combined with CT scanning in patients to provide both functional and anatomical information. [ 18 F]-FDG accumulates in metabolically active cells after phosphorylation, and so is taken up by metabolically active macrophages within the TB granuloma and other inflammatory foci. A primary limitation of PET imaging for TB is the high cost and low availability in developing world. Increased PET uptake is well described in both pulmonary and extrapulmonary TB lesions (Matsuura et al., 2000;Bakheet et al., 1998;C.M. Yang et al., 2003), and can cause diagnostic uncertainty with malignancy and other infections (Chen et al., 2004;Li et al., 2008).
PET scanning is clinically useful in certain patients with TB. For example, when patients have normal radiology but symptoms highly suggestive of active TB, PET scans may identify occult foci of infection which can then be sampled to confirm the diagnosis and for culture (Figure 7, arrow). Furthermore, PET imaging has been proposed for monitoring the resolution of TB disease (Hofmeyr et al., 2007), although the benefit must be weighed against the increased radiation exposure. Current research questions which need to be addressed are whether PET imaging can be useful to define cure, especially in the context of drug-resistant TB where treatment regimes may exceed 18 months, and also investigate whether active foci can be identified in patients with clinically "latent" disease. In addition to [ 18 F]-FDG, a wide array of radiopharmaceuticals have been developed at the preclinical level which might be applied to TB (Signore & Glaudemans, 2011). However, the potential of these in man have not yet been confirmed. For example, radiolabelling the antibiotic ciprofloxacin looked promising initially to investigate cryptic foci of infection (Britton et al., 2002), but has not entered clinical practice widely. The ability to detect a wide range of pathophysiological markers suggests that PET imaging may emerge as a powerful modality to investigate the biology of TB in man, but currently most prospective candidates require further study in model systems before clinical studies in man can be considered.

Single Photon Emission Computed Tomography (SPECT)
A number of SPECT radiotracers have been applied to the management of TB including 99m Tc-methoxyisobutylisonitrile ( 99m Tc-MIBI) (Ahmadihosseini, 2008), 67 Ga (Liu et al., 2007) and 111 In-octreotide (Vanhagen et al., 1994). 99m Tc-MIBI is a widely used myocardial perfusion agent, which can accumulate in tumours and inflammatory lesions (Aktolun et al., 1991;Caner et al., 1992;Kao et al., 1994;A. Yang et al, 2007). Ahmadihosseini and colleagues studied 36 patients with either proven active or inactive treated pulmonary TB and found that 99m Tc-MIBI uptake was increased in 23 out of 24 patients (95.8%) with active pulmonary TB but none of those with inactive TB (Ahmadihosseini et al., 2008). M. tuberculosis has been demonstrated to have significantly higher 99m Tc-MIBI uptake compared with fibroblasts and myocytes cultures (Stefanescu et al., 2007), suggesting the bacilli themselves contribute to the signal detected on 99m Tc-MIBI SPECT images.
A number of studies have found positive SPECT images in sputum smear negative patients subsequently found to have a positive sputum culture for M. tuberculosis (Ahmadihosseini et al., 2008;Önsel et al., 1996;Stefanescu et al., 2006), suggesting that SPECT imaging may be very useful while awaiting culture results. However, as previously stated, many benign and malignant etiologies can also demonstrate 99m Tc-MIBI uptake (Aktolun et al., 1991;Caner et al., 1992;Kao et al., 1994;A. Yang et al, 2007), which significantly decreases the value of this radiotracer in the differential diagnosis of pulmonary TB from other lung pathologies.

Future prospects
A fundamental challenge of TB research is to develop applications which are useful in resource poor settings and can be deployed with minimal investment in infrastructure, maintenance and staff expertise. Furthermore, any application needs to be equally applicable in urban and rural settings. Most modern imaging techniques are useful in the developed world, but are not available to the vast majority of patients with TB who live in resource-poor settings.
However, the detailed study of a small number of patients may identify pathophysiological markers of TB which can then be simplified to develop new diagnostic and therapeutic approaches applicable in resource poor settings. Ironically, when one considers molecular imaging of TB in this light, developing an imaging technique based on plain chest radiography is currently the only widely deliverable approach in the near future. For example, if a highly radio-dense specific TB ligand was developed, a diagnostic test might involve taking an initial chest x-ray, injecting the labelled ligand, and then taking a second x-ray to identify high uptake in the region of TB. This might be useful in the common clinical scenario when a patient presents with upper zone fibrosis, which may be caused by either old self-healed TB or new active TB. If the patient is sputum smear negative, an expensive and invasive bronchoalveolar lavage is required, so a non-invasive test to determine disease activity would be useful.
Another frequent clinical scenario is a patient with immunological evidence of infection, but with a normal chest x-ray and a cryptic location of disease. An investigation whereby one could locate the site of disease for aspiration and culture analysis would be clinically useful. This assay might either rely on antimycobacterial ligands, potentially using the "dock and lock" strategy (Goldenberg et al., 2007), whereby a primary antibody is first injected which docks on the mycobacterial target, and then 24 hours later a second radiolabelled antibody is injected which locks onto the primary antibody, or alternatively might focus on the host immune response, such as looking for increased metalloproteinase activity at the site of disease (Elkington et al., 2011).
In additional to diagnosis, a secondary role of imaging is to determine the prognostic and therapeutic correlates of host immunity. Currently, standard treatment lasts for six months. A recent trial comparing short-course therapy for four months in patients with low risk features was stopped because of increased recurrence in the short course treatment group (Daley, 2010). We need better markers to identify patients who will respond rapidly to treatment and imaging modalities to define mycobacterial load, the effectiveness of the host immune response and TB cure.

Conclusions
It is clear that molecular imaging technologies will play an important role in improving our understanding of the host-pathogen interactions that occur in animal models of TB, and should speed up preclinical testing of novel vaccine candidates and therapeutic regimes. In the clinical setting, standard radiographic imaging is likely to remain the mainstay of TB management in humans, with PET imaging emerging as a useful adjunct in specific cases. Indeed, molecular imaging in human TB also has the potential to provide valuable insights into disease pathogenesis, and to aid in the development of new diagnostic approaches and treatment options. However, it remains to be discovered whether approaches currently at preclinical development can be rolled out to the regions of the world where TB is pandemic. We hope that the detailed investigation of small number of patients using cutting-edge technologies may identify pathological markers which can then be developed into simpler, plain radiograph based assays which fulfil the requirements of a truly valuable TB diagnostic.