Special Viewpoints

Molecular Imaging: Looking at Problems, Seeing Solutions

See allHide authors and affiliations

Science  24 Oct 2003:
Vol. 302, Issue 5645, pp. 605-608
DOI: 10.1126/science.1090585


Noninvasive molecular-imaging technologies are providing researchers with exciting new opportunities to study small-animal models of human disease. With continued improvements in instrumentation, identification of better imaging targets by genome-based approaches, and design of better imaging probes by innovative chemistry, these technologies promise to play increasingly important roles in disease diagnosis and therapy.

Most biologists associate “imaging” with microscopic, cell-based studies. Our familiarity with noninvasive imaging procedures is primarily personal; we think in terms of our dental x-rays, the magnetic resonance imaging (MRI) necessitated by our overzealous passion for tennis, or the diagnostic positron emission tomography (PET) scan for a family member threatened with a devastating disease. This limited view may soon be a thing of the past, however. A new set of technologies, collectively termed “molecular imaging” (14), is providing biologists with exciting new opportunities to study noninvasively and repetitively the mechanisms underlying normal development and disease in small-animal models. Using these imaging technologies, researchers can now measure, in living animals, dynamic biological processes such as metabolic activity, cell proliferation, apoptosis, receptor occupancy, reporter gene expression, and antigen modulation. Here, I briefly describe several of these new technologies, provide a few examples of the ways they are being applied to preclinical research, and discuss prospects for their translation to the clinic.

Imaging Instrumentation and Probes

Molecular imaging, which I define here as the “noninvasive, quantitative, and repetitive imaging of targeted macromolecules and biological processes in living organisms,” requires two basic elements: (i) molecular probes whose concentration and/or spectral properties are altered by the specific biological process under investigation and (ii) a means by which to monitor these probes.

Imaging probes are typically light- or near-infrared (NIR)–emitting molecules or molecules labeled with radioisotopes. Two general types of probes are used in functional imaging studies, direct binding probes and indirect probes. Positron-emitting analogs of dopamine {e.g., 3-(2-[18F]fluoroethyl)spiperone ([18F]FESP)}, used to monitor the dopamine receptors of the striatum, are examples of direct binding probes. Indirect probes, by contrast, reflect the activities rather than the concentrations of their macromolecular targets. One indirect imaging probe that is widely used in neurological, cardiovascular, and oncology investigations is the hexokinase substrate 2-deoxy-2-[18F]fluoro-D-deoxyglucose (FDG), which monitors glucose metabolism. When FDG is administered systemically, it is accessible to essentially all tissues. However, it is retained in cells only if it is phosphorylated by hexokinase; FDG retention thus reflects hexokinase activity. Radionuclide-labeled probes are detected by PET or SPECT (single-photon emission tomography), probes emitting light (fluorescence, bioluminescence, or NIR emissions) are detected by optical imaging, and radiowave emissions are detected by MRI. The recent development of user-friendly instrumentation designed specifically for small-animal imaging has been a driving force for molecular-imaging studies at the preclinical level. Dedicated small-animal devices are now available for radionuclide-based imaging (e.g., microSPECT and microPET) (1), optical imaging of visible light [using sensitive, cooled charged-coupled device (CCD) cameras] (5), and NIR emissions (6).

A major turning point in molecular-imaging studies of living animals was the introduction of noninvasive reporter gene assays [reviewed in (1, 2, 5, 7)], extending to in vivo investigations a technology already in widespread use by molecular and cell biologists. Radionuclide-labeled probes have been used to monitor, in living mice, the expression of reporter genes such as ectopically expressed dopamine D2 receptor (D2R), using the direct-binding FESP probe, or the herpes simplex virus type 1–thymidine kinase (HSV1-TK) (1, 2, 7). HSV1-TK is monitored with positron-labeled thymidine analogs. Like FDG, the indirect substrate probe for hexokinase, positron-labeled substrates for HSV1-TK are retained in cells as a result of enzyme-dependent phosphorylation (1, 2, 7). For optical-imaging assays, the most commonly used reporter genes are the firefly or Renilla luciferase enzymes. The light produced by these enzymes from their substrates (luciferin or coelenterazine, respectively) is monitored with sensitive CCD cameras (5). In the future, new reporter genes encoding fusion proteins that can be imaged with fluorescent, bioluminescent, or radionuclide probes (8) will allow researchers to study a single animal with a number of different imaging probes and instrumentation appropriate for distinct applications.

Molecular Imaging in Preclinical Studies

One important advantage of molecular-imaging experiments is that each animal serves as its own “control” for previous and subsequent analyses; thus, experimental uncertainties arising from interanimal variations are greatly reduced. Consecutive analysis of the same animal also means that fewer animals are needed for each study, which is both a more cost-effective and a more ethical alternative to experiments that require sacrificing groups of animals for each time point. Molecular imaging has already been widely applied to many areas of preclinical research. Only a few illustrative examples are mentioned here; for comprehensive reviews, readers are referred to (17).

Molecular-imaging technologies have been used often by researchers studying mouse models of tumor development. In one application, Vooijs et al. (9) generated transgenic mice in which activation of luciferase expression was coupled to deletion of the tumor suppressor gene Rb (encoding the retinoblastoma susceptibility gene product). Because loss of Rb triggers the development of pituitary tumors in this model, these workers were able to monitor tumor onset, progression, and response to therapy in individual animals by repeated CCD imaging of luciferase activity.

Many drugs for cancer and other disorders exert their therapeutic effects by inducing apoptosis. The ability to repetitively image apoptotic responses in living animals will greatly facilitate preclinical evaluation of these drugs. As a step toward this goal, Laxman et al. (10) engineered a luciferase reporter molecule that remains largely inactive until it is cleaved by caspase-3, an enzyme that is produced in apoptotic cells. In xenografted mice bearing tumors stably transfected with this reporter construct, luciferase activity, detected with the CCD camera in living animals, was greatly enhanced when the mice were systemically treated with a proapoptotic drug that activates caspase-3. Transgenic mice carrying this reporter gene could also be used to monitor apoptosis induced by other agents, such as inflammatory stimuli and environmental toxins.

For researchers studying transgenic mice, identification of founder mice that express the transgene in the proper spatial and temporal pattern often requires extensive, time-consuming screening by biochemical assays of progeny from each potential founder. The ability to monitor transgene expression by noninvasive imaging in living animals permits the rapid and simple identification of founders without breeding. Rapid identification of mice expressing luciferase from the heme oxygenase and bone morphogenetic protein–4 (BMP4) promoters are examples of this application (11).

Optimization of gene-therapy protocols has been hampered by the inability to monitor the location, magnitude, and duration of expression of the therapeutic gene. Molecular imaging can provide this critical information, if the therapeutic and reporter proteins are coexpressed from the same mRNA. This principle is illustrated in Fig. 1, which shows micro-PET images of a mouse several days after injection of an adenoviral vector that encodes the D2R gene and the HSV1-TK reporter gene within a bicistronic transcription unit. Expression of the two genes in the liver (the primary site of adenovirus accumulation in mice) was found to be highly correlated over a 3-month period and over a 7- to 10-fold concentration range (12).

Fig. 1.

A replication-defective adenovirus was constructed that contains a bicistronic transcription unit encoding the D2R reporter gene and the HSV1-TK reporter gene. This virus was injected into the tail vein of a mouse. Two days after viral administration, the mouse was injected intravenously with the positron-labeled D2R ligand [18F]FESP (12) and then imaged in the microPET scanner. One day later the mouse was injected intravenously with the HSV1-TK substrate 9-(4-fluoro-3-(hydroxymethylbutyl)guanine ([18F]FHBG) and imaged again by microPET. These images reveal an excellent correlation between the expression of the two genes in the liver (the primary site of adenovirus accumulation in mice). CMV, cytomegalovirus early promoter; IRES, internal ribosomal entry site.

By coupling optical imaging with targeted gene transfer, Adams et al. (13) were able to detect occult tumor metastases in mice (Fig. 2). These workers constructed an adenovirus vector expressing luciferase from the prostate specific antigen (PSA) promoter and injected it into mice bearing human prostate tumor xenografts. Optical CCD imaging revealed luciferase activity in tissues outside of the primary tumor, including the spine. These extratumoral signals were traced to metastatic lesions that had produced no clinical symptoms in the animals.

Fig. 2.

A nonreplicating adenovirus expressing firefly luciferase from a modified PSA promoter was injected into a human prostate tumor xenograft. Optical imaging of luciferase expression is shown in the panel on the left. The primary tumor is on the lower left flank. Additional, extratumoral signals are apparent. When tissues were dissected and analyzed for luciferase expression, a portion of the spine was positive for luciferase (upper right panel), suggesting a potential metastasis. Histologic analysis of this sample (lower right panel) demonstrated the presence of metastatic tumor. [Reprinted with permission from (13). Copyright 2002 Nature Publishing Group (www.nature.com/nm/)]

Molecular imaging of reporter genes can also be used to monitor the biodistribution and efficacy of cell-based therapies. For example, imaging of tumor-targeted T cells expressing the HSV1-TK reporter gene has allowed researchers to monitor the extent and duration of tumor targeting after intravenous infusion of these cells to tumor-bearing mice (14, 15).

Moving Molecular Imaging into the Clinic

In vivo reporter gene technologies require expression of an exogenous gene in target tissues. Translation of reporter gene molecular imaging to the clinic will therefore be limited to studies of gene therapy and cell-based therapies (e.g., replacement of striatal neurons in Parkinson's disease and T cell therapy for cancer) (16).

To broaden the clinical applications of molecular imaging, additional probes must be developed to monitor endogenous targeted molecules and biological processes. Genetic, genomic, and proteomic approaches will be important tools in the search for new biomarkers that reflect disease progression and response to therapy. The imager's job is to develop specific probes to monitor the amounts (with direct-binding probes) and/or catalytic activities (with indirect probes) of these key mediators and/or indicators of endogenous processes. Among the many new therapeutic and imaging targets being explored are caspases for monitoring apoptosis, cyclin D for measuring tumor cell proliferation, mitogen-activated protein kinase (MAPK) for measuring activation of the Ras signaling pathway, and cyclooxygenase-2 for monitoring inflammation and/or cancer progression.

Direct-binding molecular probes such as antibodies and receptor ligands bind stochiometrically to their targets. For targets overexpressed in pathological conditions, such as amplified HER2/neu receptors in breast cancer, direct-binding imaging probes will be of substantial clinical benefit in monitoring both tumor burden and response to therapy. Again, genomics will play an important role in defining new direct-imaging targets: Gene expression profiling and proteomic analyses of human tumors have already begun to identify proteins whose differential expression defines subclasses of tumors with differing prognoses.

For analysis of many disease processes, the important imaging targets are in vivo measurements of pathological enzyme activation, not enzyme concentration per se. For example, the activation of MAPK, not the amount of MAPK protein, is a reflection of the increased mitogenic signaling activity often found in cancer. Similarly, the increased kinase activity of the mutated c-Kit receptor, not the number of c-Kit receptor molecules, is the important target in gastrointestinal stromal tumors. Indirect imaging probes such as FDG and luciferin report the activity of target enzymes, not their stochiometric levels. For molecular-imaging assays of a wide variety of cancers, the development of radionuclide-labeled small-molecule substrates for critical kinases, radionuclide-labeled substrates that, when phosphorylated, will remain inside the cell, is a major goal. Similarly, for optical imaging, the search is on for substrates whose spectral emission characteristics are altered as a result of the activity of target enzymes. Noteworthy progress has been made in the development of “activatable” optical-imaging probes for a wide variety of proteases. These probes are substrates whose products emit NIR fluorescent light after protease cleavage (17).

Once molecular targets are defined by genomics and validated, new target-specific imaging probes must be developed. The problems encountered in designing direct and indirect molecular-imaging probes are similar to those encountered in designing new drugs, although the criteria for imaging probes are often more stringent. Imaging-probe chemistry has traditionally been oriented toward synthesis of labeled drug analogs. To move forward in the clinic, molecular imaging will require more innovative chemistry. High-throughput screening of chemical libraries and generation of libraries that cover extensive chemical space are expected to play key roles. Large recombinant antibody and phage libraries will also provide new direct-binding molecular-imaging probes for target molecules. There is likely to be a substantial increase in commercial development of molecular-imaging probes, both radiopharmacy production for PET and SPECT imaging and conventional chemistry for optical probes.

Which functional imaging technologies will make the most headway in new clinical applications? Some researchers are betting that radionuclide techniques will continue to lead the way, as they have done particularly in the neurosciences (1820), whereas others feel that NIR, with a potential depth of 10 cm, will come to the fore (6). Because of their substantial attenuation, bioluminescent probes are not likely to translate extensively into clinical use. New target-specific molecular-imaging agents, such as radionuclide-labeled probes (for PET and SPECT) and optical-imaging probes (for fluorescent, bioluminescent, and NIR assays), must go through clinical trials for Food and Drug Administration approval. In general, nanogram levels of radioactive probes are required for radionuclide analyses, whereas microgram or milligram probe concentrations are often required for optical imaging. Because radionuclide probes require substantially less mass, they may be incorporated into clinical practice more readily than probes that are required in greater concentrations.

The Future of Molecular Imaging

At the preclinical level, molecular imaging has clearly arrived. The National Cancer Institute has already established seven academic In Vivo Cellular and Molecular Imaging Centers (ICMICs), and more are planned. A new National Institutes of Health institute emphasizing molecular imaging has been created. Two dedicated societies, two journals, and three annual meetings have been founded. Cooperation between cancer centers and molecular-imaging programs has fostered, in many institutions, the development of dedicated small-animal imaging shared resources. To be effective, these facilities must provide transparency, easy access, robustness, reproducibility, convenience, and intellectual guidance. Online scheduling, technical assistance and advice, a consumer-oriented business plan, online data retrieval, and standardized operating procedures should lower the activation energy for potential users, demystify the molecular-imaging technologies, and reduce the procedures to good laboratory practice.

For small-animal imaging, more sophisticated technologies are just around the corner. For example, 1-μl volumetric resolution microPET instrumentation will provide better anatomic discrimination of functional assays: pinpointing the locations of tumors within organs, determining the location of cell migration more accurately, etc. (21). Fluorescence-mediated tomography (6) will improve the resolution and quantitation of optical-imaging procedures. Spectral-imaging technologies (22) will discriminate emissions from multiple fluorescent probes, permitting simultaneous analysis of distinct optical probes and dramatically reducing background autofluorescence.

In the clinic, new applications of conventional imaging technologies are likely to play increasingly important roles, particularly in oncology. Several new cancer drugs under development (e.g., certain antiangiogenic compounds) are cytostatic agents that arrest tumor growth but do not necessarily reduce tumor bulk. Thus, tests that monitor changes in tumor volume, the traditional criterion used to evaluate therapeutic efficacy, may be inadequate. Measurements of tumor cell metabolism with FDG and tumor cell proliferation with 3′-deoxy-3′-[18F]fluorothymidine may serve as interim evaluative procedures until target-specific imaging probes for caspases, activated kinases, angiogenic factors, and other specific biomarkers are developed.

Molecular imaging is expected to play a key role in drug discovery, development, and delivery at the preclinical level (23). Importantly, it will not be necessary to develop imaging analogs of every drug candidate; in vivo competition assays of unlabeled compounds with labeled probes for agents with known pharmacological characteristics and efficacy can be used in the drug evaluation process. Noninvasive characterization of drug targeting, receptor occupancy, concentrations required for effective receptor or enzyme inhibition, etc., without the requirement to sacrifice large numbers of animals, will speed the evaluation of lead compounds. As new drug candidates proceed through pharmacodynamic and pharmacokinetic studies, imaging analyses can quickly, quantitatively, and repetitively monitor target accessibility, duration of retention at the target site and its correlation with drug efficacy, and clearance from irrelevant tissues. Once clinical trials are initiated, imaging assays that monitor the location and concentration of the experimental drugs will facilitate evaluation of both their pharmacological properties and their therapeutic effectiveness in patients.

Finally, by combining new imaging probes with multimodality-imaging instruments that merge structural and functional data, physicians should be able to perform multiple functional-imaging assays simultaneously with anatomic analyses. Information derived from structural studies and from noninvasive, repetitive monitoring of drug distribution and concentration can then be correlated with biological effects on signal transduction pathways, target enzyme activities, antigen levels, receptor activation, cell proliferation, proteasome activity, etc. These noninvasive assays will permit real-time monitoring and modification of targeted interventions and therapeutic strategies. As it realizes its potential, “molecular imaging,” like “molecular biology,” will cease to be thought of as its own discipline and will become standard practice in research and medicine.

References and Notes

View Abstract

Navigate This Article