Cardiovascular Metabolic Imaging: Physiologic and Biochemical Dynamics In Vivo

Abstract

Fifty-eight investigators from the fields of biochemistry, physiology, cardiology, nuclear medicine, and physics met to discuss the development of metabolic imaging techniques for application to cardiovascular and pulmonary studies in health and disease. The workshop was sponsored by the Division of Heart and Vascular Diseases, National Heart, Lung, and Blood Institute and was held on September 16 to 18 in Bethesda, Maryland, in facilities provided by the American College of Cardiology. This report summarizes the presentations and discussions and presents recommendations for future studies.

Introduction

The purpose of the conference was to bring together basic scientists and clinical investigators working in cellular physiology, biochemistry, noninvasive imaging, and clinical cardiopulmonary research to define new goals and directions in the field of metabolic imaging. Metabolic imaging involves the utilization of positron-emission tomography (PET), nuclear magnetic resonance (NMR), or other (x-ray CT, ultrasound) technology for obtaining dynamic and steady-state information on the uptake, metabolism and fate of substrates for cellular metabolism, and for hormones or drugs acting at receptor sites. The new technologies have promise in providing quantitative data, in vivo, allowing refinements in the interpretation of high resolution spatial images in terms of rates of transport across membranes, intracellular reactions and cell functions. Keys to image interpretation and to understanding the limitations are kinetic modeling of the physiologic and biochemical events and knowledge of the biochemical changes occurring in pathophysiologic states.

Recent years have seen several new developments that bear on experimental design and image interpretation. These include: (1) rapid improvements in NMR technology, providing better chemical resolution and a potential for imaging more than protons; (2) improved temporal and spatial resolution for PET; (3) rapid advances in kinetics of mass transport and transmembrane exchange; (4) deeper insight into biochemical regulation and control, particularly for living functioning cells and tissues as components of intact organs; and (5) the recognition of the physiologic and biochemical importance of the endothelial cell as a regulator of vasomotion and of solute exchange. All this is occurring at a time when new capabilities for reopening blocked arteries are stimulating studies on reperfusion injury, particularly to the heart.

The main topics addressed were substrate biochemistry in the normal heart, the approaches to experimental design and modeling analysis as might be used with both PET and NMR, and the biochemical shifts occurring in pathophysiologic states. The final session was a discussion of future directions and included possible studies in which both PET and NMR might be used in collaborative rather than competitive fashion.

I. Metabolism: overview of current dilemmas and future directions

The opening session of the workshop concentrated on control pathways of metabolism, including carbohydrate, lipid, and nitrogen metabolism as well as the use of NMR and PET to trace metabolic flow in vivo. The objectives of this session were: (1) to set the background in cardiac metabolism, (2) to pose questions that may be answered by NMR/PET, and (3) to provide impetus for future directions of tracer imaging in vivo.

Carbohydrate metabolism

Regulation of carbohydrate metabolism involves an intricate system of enzymatic pathways, which coordinates myocardial energy production from glucose and glycogen. Control mechanisms in glycolysis (phosphofructokinase and glyceraldehyde-3-phosphate dehydrogenase) were discussed with reference to substrate cycling events, the importance of the latter being unknown in heart. Regulation of glycogen synthesis and degradation takes place by four main mechanisms, i.e., allosteric control, covalent modification (coordination of synthetic and degradation pathways), hormonal control through the cAMP/protein kinase cascades, and Ca²⁺ activation of phosphorylase kinase. The end product of glycogenolysis and aerobic glycolysis is pyruvate, which is transported into the mitochondria and oxidatively decarboxylated by pyruvate dehydrogenase; this reaction is subject to control by feedback inhibition and phosphorylation-dephosphorylation events. Subsequent control of pyruvate oxidation as acetyl-CoA in the tricarboxylic acid cycle takes place at citrate synthase, isocitrate dehydrogenase and α-ketoglutarate dehydrogenase. Transport of reducing equivalents from the cytosol to the mitochondria takes place by substrate shuttles, with the malate-aspartate shuttle predominating over the α-glycerophosphate shuttle in the heart. Finally, selection of fuel by the heart (glucose vs free fatty acids) may rest with regulatory enzymes that are controlled by factors other than substrate availability and product removal.

Lipid metabolism

Pathways governing fatty acid utilization in the heart were discussed. Fatty acid uptake from either fatty acid: albumin complexes or lipoprotein triglyceride across the endothelial and sarcolemmal membrane is believed to occur by passive diffusion, although evidence is growing for carrier-mediated transport. Fatty acids are activated on the mitochondrial outer membrane; there is potentially a physiologic role for fatty acid binding protein in cellular fatty acid buffering in heart but this is unproven. The extent to which fatty acyl-CoA levels are controlled by acyl-CoA hydrolase also is not known. Fatty acid oxidation by chain-shortening of very long acyl-CoA molecules occurs by peroxisomal β-oxidation, primarily in liver; similar documentation for this pathway in heart is not available. Mitochondrial β-oxidation may be controlled by fatty acid availability, flavin-linked dehydrogenase, disposal of acetyl-CoA via regulation at the ATP/ADP ratio, and disposal of reducing equivalents. Recent evidence suggests control of acyl-CoA conversion to acylcarnitine at the carnitine palmitoyl-CoA transferase-I on the outer aspect of the inner mitochondrial membrane. This reaction is apparently regulated by malonyl-CoA, which has been found to be present in heart, by alterations in physiologic state and by pH. A 90,000 molecular weight protein may confer inhibitor sensitivity to the transferase. The final category discussed in fatty acid utilization is ketone body oxidation. Since the heart does not generate ketones or export citrate from the mitochondrial matrix, a high acetyl-CoA “pressure” exists in heart mitochondria. Some reduction in acetyl-CoA concentration can occur with acetylcarnitine synthesis and export to the cytosol.

Nitrogen metabolism

Nitrogen metabolism in the heart was presented with respect to control mechanisms of gene expression. Control of gene expression occurs at transcription that is measured by messenger-RNA (mRNA) synthesis, and at the posttranscription level which involves processing of the mRNA for delivery to the cytosol as well as storage and degradation of the gene message. Transcriptional activity is measured as RNA synthesis, which is activated during periods of increased growth. Increased RNA synthesis is observed by measurement of RNA levels, by pulse labeling of RNA with its nucleotide precursor (reutilization of mRNA precursor is limited in the heart), or by determination of DNA-dependent RNA polymerase activity. Recombinant DNA technology allows measurement of mRNA coding for specific proteins. Correlation between the amount of transcribed mRNA and the synthesis of the corresponding protein suggests that transcription is the major mechanism of regulation in growth.

The second level of potential control is posttranscriptional, i.e., regulation of the number of functionally competent mRNA molecules available for translation. Steps involved in posttranscriptional control are mRNA processing, transport of mRNA to the cytoplasm, and degradation of the message.

Translational regulation of gene expression involves the ribosomal cycle, including initiation of protein synthesis, elongation, and termination. Factors regulating these processes are the availability of energy, the degree of phosphorylation of regulatory proteins, and the state of the ribosomal subunits. The final level of control is degradation of the gene product.

Protein turnover in heart is difficult to quantitate, due to recycling of radiolabeled amino acids, changing the relative concentrations of synthesized products. Thus turnover cannot be accurately assessed using total cell protein, amino acid synthesis, and degradation rates. For example, to study isoproteins (proteins similar in function yet products of different genes) a single molecular variate must be selected and isolated.

The current challenge in studies of gene regulation and expression includes analysis of DNA organization and gene structure. The significance of introns (silent, intervening nucleotide sequences) to the expressed nucleotide sequence (exons) is unknown. The process by which exon transcription is derepressed and made available to the action of RNA polymerase is also obscure.

Metabolic assessment by NMR

Application of in vivo imaging and spectroscopy to the study of cardiac energy metabolism is motivated by the ability to probe the biochemical state of the heart noninvasively and repeatedly. Development of proton imaging for clinical use has accelerated biochemical measurements using other nuclei, e.g., ³¹P, ¹³C, ²³Na, and ¹⁹F.

Cardiac studies concerning ³¹P in isolated, perfused hearts have dealt mainly with global rather than regional metabolism. In vivo information on ³¹P-metabolite levels demonstrates that concentrations of intracellular phosphate are lower than those obtained by freeze-clamp techniques, and that ADP levels measured by conventional techniques seriously overestimate free ADP levels. The concentration of ADP and P_i in the cell are important in determination of the free energy of ATP hydrolysis in vivo.

Saturation transfer techniques have been used in NMR studies to determine the enzymatic exchange between creatine phosphate and ATP by creatine kinase. Exchange fluxes are found to be three to five times the oxygen consumption, indicating that creatine phosphate is a buffer for rapid changes in ATP levels. Although the creatine kinase shuttle has been suggested to act to transport energy-rich compounds from the mitochondria to the myofibrils, diffusion calculations demonstrate that flux due to ATP diffusion alone is sufficient to explain oxygen consumption in the absence of creatine phosphate. However, in the normal situation where creatine concentrations in the cell are high compared to ADP, the preponderance of the flux is carried by creatine kinase. Studies using NMR can be designed to answer the question whether creatine kinase is obligatory to energy, translocation.

NMR spectroscopy can report cellular pH by inorganic phosphate shifts. The intracellular concentration of free Mg²⁺, important as an effector of metabolism, can be estimated by the position of the phosphorous resonances of ATP. Independent estimates of free Mg²⁺ concentrations can be derived from ¹³C measurement of citrate resonance positions.

^I3C NMR holds promise for elucidating control mechanisms in carbohydrate and fatty-acid metabolism. By sequential sampling following a pulse label of ¹³C, the kinetic evolution of a metabolic conversion can be fitted to the time course of label appearance in various metabolites. ¹³C NMR may also be useful in improving the sensitivity of proton signals attached to the ¹³C labeled atom of various metabolites. The high proton abundance coupled to improved detection may be potentially important in future development of in vivo spectroscopy.

Metabolic assessment by PET

The utility of PET in metabolic studies applies to sensitivity measurement, and quantitative delineation of metabolic pathways by this technique is not possible at present. Agents employed for metabolic imaging are physiologic substrates, which are metabolized to their labeled products, and substrate analogues, which are refractory to further metabolic conversions and are trapped in the tissue.

The spatial distribution of ¹¹C-palmitate can be used to delineate qualitative metabolic abnormalities associated with clinical disease. While the initial regional delivery of' ¹¹C-palmitate as observed in vivo by PET is proportional to flow, and a diminution in initial accumulation in ischemic regions reflects this, there is also diminished utilization due to decreased aerobic metabolism, which results in more back flux (and shorter retention) of unmetabolized fatty acid than occurs in nonischemic regions. Parallel studies using ¹⁴C-palmitate in animals were confirmatory: there was an increase in ¹⁴C-palmitate in lipid pools and decreases in ¹⁴C-labeled metabolites in tissue and in released ¹⁴CO₂.

In isolated hearts, the extraction of ¹¹C-glucose and of ¹⁸F-deoxyglucose increase in the initial phases of ischemia, but decrease with prolonged ischemia. While transient increases in regional anaerobic glycolysis result in transiently increased uptake rates, infarcted regions exhibit decreased uptake.

Substrate imaging by PET is clinically useful in evaluating the response of the ischemic myocardium to interventions, e.g., changes in flow or metabolic processes following coronary thrombolysis. Despite limitations of PET for quantitation of metabolic flux, tracers have been useful in assessment of perfusion and metabolic responses to interventions. Other examples of metabolic images used to characterize disease states clinically are spatially heterogeneous tracer accumulation in cardiomyopathic heart in vivo and the differentiation of zones in the heart by metabolic tracers delineating the progress and extent of myocardial infarction.

The use of NMR/PET technologies to evaluate metabolism in vivo can open a new chapter in metabolic control by combining the strengths of each technique to obtain biochemical information related to a defined physiologic condition. This type of information can provide clues to the metabolic etiology of disease rather than defining the end state of pathology. Implementation of in vivo imaging techniques requires a diversity of knowledge uncommon to biological research. However, these techniques potentially provide answers to complex questions not otherwise approachable in vivo.

II. Metabolism in myocardial disease

Three areas of particular interest are decreased blood flow, increased cardiac load, and cardiomyopathy. Blood flow reductions that are great enough to be labeled “ischemia” tend to be associated with considerable regional heterogeneity of flow. While transmural gradations in flow are amongst the most clinically useful observations, imaging techniques have not provided sufficient spatial resolution to assess them under physiologic or clinical conditions. The ability to identify ischemia is obviously seriously compromised in the situation where a subendocardial ischemia is accompanied by a reflexly mediated subepicardial hyperemia. Thus high-resolution data on regional flows is a prerequisite to deeper understanding of local metabolic events.

There is a provocative hypothesis that the myocardial response to chronic pressure overload involves processes similar to those in idiopathic cardiomyopathy. While the relationship between overload-induced hypertrophy and intracellular biochemical changes with diminished tension development are not understood in terms of the underlying mechanisms, the association merits consideration. The reversibility of overload hypertrophy is only partial, again raising the question of mechanism. The ischemia scenario, with metabolic needs exceeding supply, resulting in diminished contractile function, segmental necrosis, and localized fibrosis is in some ways analogous to the events induced by pressure overload or by cardiomyopathy. Diabetic cardiomyopathy and hypertension, in this framework, may also be related, as in the combination of renal hypertension and diabetes leading to cardiomyopathy.

Likewise, transient ischemia, which is insufficient to cause an abnormality of cell membrane function may cause abnormalities of intracellular metabolism. In contrast, cell membrane dysfunction allowing extra calcium entry may well be related to focal arterial spasm, followed by regional ischemia. In the latter case the imaging of cellular metabolism is of less interest than is the treatment of the vascular hypercontractility. The incentives are therefore strong to obtain quantitative information not only of the intracellular metabolic events, but also of the abnormalities of local perfusion, including sites of arterial obstruction. Well defined models that include both structure and metabolism are needed in the quest for posing the investigative questions clearly.

Membrane structure and function are now coming into focus as being of fundamental importance in the pathogenesis of disease. Events altering membrane structure change intracellular function. Drugs or natural compounds that accumulate in membranes can affect the transport processes, for example, as ethanol affects the calcium transport ATPase. Chronic ethanol feeding does more; in animals it alters the fatty acyl chain composition of the membrane phospholipid bi-layer increasing the level of saturation of fatty acids. This in turn may change membrane fluidity and affect membrane transporters or receptors.

Normal substances in abnormally high concentrations also affect membranes; high fatty acid levels during ischemia have detergent effects. Micelles may form with phospholipid dislocation, membrane disruption, and cell death. Related phenomena occur with lipolysis, during which arachidonates and other substances may be released and initiate complex messenger signaling processes. Together with or separately from the effects of lipolysis are the effects of free radicals, a probable contributor of reperfusion injury. Not to be neglected amongst initiators of membrane dysfunction are viruses (e.g., hepatitis B virus) producing antigenic proteins that activate immunologic defenses and affect membrane function. A prime target of in vivo imaging might then be discernment of membrane processes in vivo rather than simply intracellular biochemical events.

Of potential great importance is the use of NMR and PET imaging in guiding patient care, not just in making diagnoses but also in estimating the severity and extent of disorder, providing prognostic information, and monitoring therapy. Nonimaging techniques may give complementary views; for example, in vivo NMR spectroscopy has been used to obtain a measure of the biochemical impact of an observed oxygen utilization rate or ATPase activity by providing the intracellular ratio of phosphocreatine to free phosphate.

Similarly elegant approaches to relating function to biochemistry have employed in vivo NMR to look at the effects of hypoxemia in peripheral vascular disease, and for seeking “signposts” along the pathway to cardiac failure in neonates and animals. Observations by NMR spectroscopy of limb muscles during graded exercise showed in one study that the relationship between workload and the phosphocreatine/free phosphate ratio (P_Cr/P_i) was describable by first order Michaelis-Menten kinetics, the ratio falling with increasing workload. With vascular disease or cardiac disorder the ratio fell precipitously at low work levels. Another study used pericardiac NMR coils in hypoxic dogs; in the control state increased coronary flow apparently permitted maintaining normal ratios, but an increase in work caused a rapid fall in the P_Cr/P_i ratio and in cardiac arrest. Reoxygenation led to recovery in ratio and function, but a second period of hypoxia caused a more rapid, deeper decline in P_Cr/P_i ratio and irreversible arrest, presumably due to lactic acidosis and inhibition of oxidative metabolism. These events are similar to those in a neonate with cardiomyopathy; a low P_Cr/P_i ratio was observed with cardiac failure, and was followed by further clinical deterioration and death. The cardiac failure was also accompanied by NMR-detectable biochemical abnormalities in the brain. The thesis is that abnormally low values of P_Cr/P_i ratios give quantitative measures of disorder and are of predictive value in critical situations.

The potential for NMR imaging for obtaining direct metabolic information in clinical situations is more in the future, but useful indirect measures can be made with current technology. Proton images acquired with cardiac gating can provide such high resolution that wall thickness, wall motion, myocardial mass and ventricular volumes can be assessed with good accuracy. It is beginning to be feasible to estimate proximal coronary artery stenosis, but three-dimensional imaging is the key to accuracy and to observing vessels smaller than the main coronary vessels. The measurement of regional myocardial flows by NMR is conceivable, and has been demonstrated in dogs with local ischemia, using paramagnetic substances for image enhancement. Much technical development must precede estimation of transmural distributions of flow. Imaging high-energy phosphates or pH from proton spectra can only be done now with several centimeter resolution, but developments are occurring rapidly that give some reason for optimism, and would allow correlation with mechanical events. Another future possibility is the imaging of tissue characteristics such as water and lipid content, which may help in distinguishing normal tissue from viable ischemic and dead tissue. The mechanical consequences of infarction, e.g., valvular regurgitation from ruptured papillary muscle, are detectable. Future needs for diagnosis and for the assessment of tissue status go far beyond present capabilities and are an incentive for much effort in the development of the technology.

Likewise, PET technology, in each of its several aspects, needs further development. While the instrumentation is now approaching the ultimate spatial resolution, which is the range of the positron before annihilation and production of the directed pair of gamma photons, improvements can yet be made in image reconstruction, in tracer production, in interpretation of the kinetics for the estimation of flows, transport rates, and metabolic fluxes in spatially refined domains.

Of over 100 compounds labeled with short-lived positron emittors, only a few have been well characterized in terms of their physiologic biochemistry and kinetics. Fluorodeoxyglucose, labeled with ¹⁸F, is one of the more extensively studied; via interpretive models its uptake into tissues provides estimates of the rate of entry into cells and its phosphorylation by hexokinase. Its use has demonstrated an increase in myocardial glucose uptake in mild to moderate ischemia (and presumably due to the shift to glycolysis). The 1 to 2 cm spatial resolution was good enough to demonstrate that regions of myocardium with low flow and abnormal motion show increased uptake, which occurs with the shift from fatty acid to glucose utilization. A practical application resulting from observation of this biochemical shift might be in deciding when to perform vascular grafting: regions with low flow and low uptake are not improved by revascularizing grafts, whereas low flow, high uptake regions have been shown to redevelop normal cardiac motion. Similar approaches may be of value in acute myocardial infarction, for in about half of infarcts, with or without Q waves, contractile function improves spontaneously, but deteriorates in the other half. The development of predictive criteria, e.g. local flow and glucose uptake, would allow selective intervention.

The role of imaging in the cardiomyopathies, global and segmental, is ill-defined as yet. The promise for finding abnormal patterns of substrate metabolism may stimulate the use of PET for the formulation of testable hypotheses, and subsequently for their testing. Such a reversal of the usual scientific approach can prove useful when so little is known, and the validity of animal models unproven.

Regional oxygen metabolism has been widely explored with PET. Currently, there are limitations to determining oxygen utilization after single bolus injections of ¹⁵O, which has prompted the use of steady-state constant inhalation techniques. Using a priori information on flow from ¹⁵O water data and on local blood volume from ¹¹C–carbon monoxide–hemoglobin data, estimates of regional oxygen utilization are calculated from flow times local extraction. Cerebral grey and white matter are readily distinguishable by this approach, which has promise for the heart when the spatial resolution and gating are good enough to show transmural profiles. Complementary studies by PET and NMR are desired to show the degree of coupling between oxygen utilization and cardiac energetics.

Too little work has been done on the lung. The pioneering work has included studies of regional ventilation-perfusion ratios and of regional water content. Sarcoid lung, in the active phase, shows increased uptake of fluorodeoxyglucose. There are great possibilities for using radioligands to ascertain the activity of endothelial receptors.

Cardiac receptors have been studied using radioligands for α- and β-adrenergic sites, and for muscarinic cholinergic sites. These all show saturable binding, stereo-selectivity, and regional variation. While one needs high-affinity radioligands to compete for the sites with native agonists, the other side of the coin is that extremely high-affinity ligands, for which there is no release during the period of observation, give a regional distribution which is purely a function of flow and not of receptor abundance. The in vivo and in vitro distributions should be carefully compared, since in vitro one can use highly controlled preparation with purified radioligands and hopefully eliminate or account for transport limitations. Examples of in vivo studies include iodinated quinuclidinyl benzilate (QNB) as a muscarinic receptor agonist in the brain; as expected, it accumulates in regions having a high density of muscarinic receptors, cortex, putamen, and caudate nucleus. PET studies using carbon-11 methyl-QNB have been accomplished in humans; the myocardial uptake appeared related to vagal tone, and the compound could be displaced by atropine. β-Adrenergic receptors in the heart have been explored with ¹¹C practolol. Although the ratio of specific to nonspecific binding is still low, better radioligands may be developed with higher affinity and lower lipophilicity. Such data do not provide a measure of receptor activity, turnover or recycling, but do give an indication of site availability and affinity. How these radioligands might be used to estimate the release of native ligands has not been worked out. What is observed when the radioligand has similar affinity to the native ligands is regional deposition governed by receptor abundance and access in the face of local concentrations of competing native agonists or inhibitors.

While the discussions tended to focus on the more newly recognized possibilities concerning membrane and receptor functions, cellular functions, mainly biochemical, remained central. Combinations of possibilities emerged. Amongst these were: (1) the leakage of ⁸²Rb from cells as a measure of membrane integrity; (2) studies of the role of endothelial cells and their biochemical transport, and regulatory functions, e.g. ¹¹C serotonin for pulmonary cells; (3) the use of ¹⁵O to assess cell viability in heart, as has been initiated in brain studies; and (4) the combined assessment of myocardial failure with PET with studies of peripheral muscle metabolism with NMR.

Major investigative groups should attempt to combine, compare and contrast PET and NMR during the next decade; this is critical to determining the best roles for each in practical clinical situations. While PET currently has the advantages of being able to detect and localize picomolar amounts of tracer, and of providing data with good temporal resolution allowing estimation of kinetics of translocation and transformation, NMR proton imaging has spatial resolution that will not be achievable with PET, even though high resolution NMR is not available for particular molecules or tracer atoms as yet. An obvious opportunity is to use NMR proton images to define the anatomic sources for PET in image reconstruction. For both, developments are needed in modeling of substrate kinetics, and the design and characterizing of specific probing molecules for individual purposes. The heart and brain have been the most studied, therefore encouraging one to explore their important interrelationships, but simultaneously bringing to mind the fact that other organ systems may benefit just as greatly from explorations via PET and NMR.

III. Kinetics of substrate transport and reaction

The processes of delivery by flow, transmembrane transport, and interstitial diffusion are prerequisite to intracellular metabolic reaction. The physiologic and biochemical processes within the cell can be expressed as experimentally testable hypotheses when these are based upon a framework of quantitative considerations of the structural anatomy of the organ and of the sets of available biochemical anatomy, that is the set of recognized available reactions. Modern computational techniques allow expression of relatively fully developed hypotheses so that the need for immense simplification is reduced. Numerical algorithms have given solutions a million times faster than analytical solutions of three-region capillary tissue models and promise greater improvements for more complex models. This increased computational capability allows exploration of the utility and adequacy of simplifications.

Structuring the sequences of events undergone by a tracer demands recognition of individual anatomic and physiologic features. Endothelial cell transport and accumulation are examples of ordinarily neglected features that can dominate the kinetics. Another is the two-sided nature of membrane transporters, where transport rates are influenced by both the cis- and trans-concentrations of substrates, competitors, and inhibitors. A third previously unappreciated important feature of capillary-tissue exchange is the axially-distributed nature of the events, the gradients in concentrations along the length of the capillary, and therefore the need for recognizing that the transport or reaction rates are not necessarily the same at the arteriolar as at the venular end of the capillary-tissue unit. Such considerations are all at variance with standard compartmental analysis.

As a particular example, studies on tracer adenosine uptake and accumulation in the heart proved especially provoking. When tracer adenosine was infused into the coronary artery inflow, at the end of a 30 min period about 90% of the retained tracer was found in the endothelial cells. This provokes recognition that (1) the heart has more than one cell type; (2) observed phenomena may be attributed to the wrong cell type unless explicit information is sought on each; and (3) experiments of several different types are required to obtain a totally logical, meaningful, and self-consistent picture of the phenomena involved. The principle is that a model, to be considered seriously as representing the biochemical-physiologic situation, must be adequate to describe diverse types of observations. These include the anatomic arrangements, the composition of the regions and their volumes, the observed rates of transport across each of the barriers involved, the assessment of the diffusional delays, the use of appropriate input forms and reference solutes when using tracers, the identification of the products of reactions, and identification, if possible, of the sites of formation of metabolites. In vivo and in vitro observations of the biochemical events are often dissimilar; PET and NMR can therefore be expected to provide new insight into the biochemistry of normal tissues.

The modeling process begins with observation and description. The physiologic system can usually be represented so as to raise the question of causality, which in turn provokes the development of an hypothesis. A fully expressed hypothesis normally provides predictions as well as explanations, and thereby leads to defining successively more stringent tests of the hypothesized causal relationships.

Making the assumptions explicit is a mechanism to force examination of the reality of each element of the hypothesis. Is each assumption quantitatively valid in the proposed setting? Fundamentally important checks are provided by conservation of mass, of solutes (substrates and metabolites) and solvents, of energy and of tracer.

An example of a well-formulated hypothesis is one in which explicitly measured parameter values can be used without adjustment in the model hypothesis, giving rise to a predicted result that matches an observed experimental result. Such a success does not prove the hypothesis correct but supports the underlying notions of causality and provides a framework upon which the design for other experiments can be explored.

Heterogeneity of local flows and other features should be accounted for in analyzing the situation. The effect of any heterogeneity of properties in any system will lead to a systematic error in the evaluation of features of the system if the analysis is based on the assumption of uniform properties. The reason for this is that the composite behavior is normally not a linear average of the individual behaviors, that is, that the average behavior of the composite whole is not explained by the average of the concentrations or of the average rates of reaction within each region of the organ.

Mathematical developments of descriptors of the heterogeneous system permit evaluation of the error in parameter estimation (for conductance, volumes, or fluxes) stemming from the assumption of homogeneity. Such efforts are useful and important since homogeneous models are in common use, mainly for reasons of simplicity in calculation, but often also because there is lack of information with respect to heterogeneity. In the absence of specific information, empirical but logical corrections can be used to reduce the error, the goal being to transform a systemic and biased error to a smaller, symmetrically random error.

Data acquisition is the primary event in the process of evaluating an idea. The ability to discriminate between hypotheses, or to parameterize the description of the system, is dependent upon the fidelity of the data and the relevance of the data to the evaluation of the hypothesis. Is each feature of the hypothesis sensitive to data acquired in a specific experiment? Using models for designing the experiments can assure the answer.

When the data are residue function curves for regional tracer content then the knowledge of the arterial input function is critical to precise delineation of the transport and exchange processes taking place in the tissue. The arterial concentration-time curve can be recorded via PET imaging of the blood pool in the left ventricular cavity; an important prerequisite is that this input function be close enough to a sharp pulse that its frequency content is sufficiently high to elicit the information on the rapid transport processes in the exchange region. This requires, for its detection and recording, high temporal as well as spatial sampling resolution. Verification by invasive sampling is important (for example, arterial sampling) for at least a subset of the studies.

In undertaking the model analysis it is important that there be a one-on-one identification of model features with physiologic features, for example, to identify anatomically and physiologically each of the barriers involved in the transport processes. A particular concern in PET is identification of the fate of the label. Since observations of regional tracer content give no information on the chemical form, such identification must take place by other means, either kinetic analysis, or chemical analyses of tissue samples. Since kinetic analysis must be subjected to verification, this reinforces the principle of acquiring several different types of experimental data on the same system.

Model simplification by reducing complexity is an important goal. There are risks in succumbing to the natural tendency to use the most simple model. The question of acceptability in the simplicity is a quantitative one that can usually be evaluated. One of the mechanisms for doing so is to build a comprehensive model that is sufficiently complex and realistic to describe a wide variety of relevant experiments, and then to test the simpler models against it. In specific cases, albeit rather limited ones, compartmental models can give adequate representation; one example is that when barrier permeability is extremely low then the representation of the region to which there is limited access as a first order mixing chamber is not bad. On the other hand when the conductances for fluxes into specified physiologically defined regions are moderate or high, then the compartmental representation fails and axially distributed systems, often with concentration-dependent conductances, need to be considered.

Noise in the data affects parameter estimation. One approach to the evaluation of the effects of noise is provided via computer routines that optimize the fit of the models to the data and provide estimates of confidence limits based on theory of least squares. An alternative “brute force” approach is to produce a model solution with known parameters, to add noise to the idealized “data” obtained in the form of model solutions, and to fit these “data” with the model using automated optimization routines to obtain estimates of the parameter values. On repeating this operation many times, a distribution of the estimates of the parameters can be obtained for each level of noise. These distributions of estimates can be compared with the true value and assessed from the point of view of whether the distribution of estimates is distributed randomly or in some skewed fashion, which is usually the case. Gaussian-based theory often underestimates the errors, and cannot show skewness.

An important attribute of positron imaging tomography is that the time sequence of signals provides estimates of tracer content as a function of time in a chosen region in one animal. This contrasts with the use of quantitative autoradiography (QAR), which requires relatively large numbers of animals sacrificed at a succession of times in order to obtain the same kind of data. Since animals differ, the accuracy of QAR is inevitably greatly reduced compared with observations of single animals as a continuous function of time. Thus even noisy data obtainable via positron tomography may have immense advantages over the alternative types of experiments.

The general viewpoint is that good modeling analysis is based on fundamental principles of conservation of material and self-consistency of the concepts involved. Examination of the models by qualitatively different experimental approaches is essential to their testing; ultimately each model will prove to be incorrect or incomplete, leading to new hypotheses, which are then to be tested by yet more refined experiments. The techniques and approaches are applicable to a wide variety of organs and tissues, not simply to studies involving the prime substrates, but also to regulatory substances, hormones, and drugs. The recommendation is that major efforts be made in the methods of quantitative analysis; the most general models should be testable by data from both PET and NMR. Realistic, mass conservative models, tested by carefully acquired data, need to be developed as a means for analyzing the new imaging data meaningfully.

IV. Influences of flows and transmembrane transport on tracer uptake and retention by emission tomography

The basic mechanism underlying residue function or uptake curves observable by PET is active or passive transport through resistance barriers (capillary and cell membrane) with a subsequent pooling or trapping of the tracer after it passes the barriers. In order for the tracer to accumulate in the tissue it must be delivered by flow; the total delivered is the time-integrated product of flow times concentration.

The simplest form of transmembrane transport is passive diffusion. The fraction of the amount in the capillary pool transported into the extravascular space per unit time is an integral, over the capillary length, of the conductance, a permeability surface area product, PS_Cap ml g⁻¹ min⁻¹, divided by the capillary volume, V_cap. Thus the rate constant for unidirectional uptake at each location is PS_Cap/V_cap. When the flow is infinitely high, then there is no diminution in concentration along the capillary length and the uptake flux per gram of tissue is the arterial concentration C_A times the rate constant, normalized by the capillary blood volume per unit mass of tissue, where ρ_tiss is tissue specific gravity:

$$\text{Unidirectional uptake},\text{ml}{\text{g}}^{-1}{min}^{-1}={\mathrm{C}}_{\mathrm{A}}\cdot \frac{\mathrm{P}{\mathrm{S}}_{\text{Cap}}}{{\text{V}}_{\text{cap}}}\cdot \frac{{\mathrm{V}}_{\mathrm{\text{cap}}}}{{\mathrm{\rho }}_{\text{tiss}}{\text{V}}_{\text{tiss}}}$$

This is the case of pure diffusion or barrier-limitation.

But since the amount present to be transported does normally depend on flow, one would expect flow to play a predominant role in kinetic analyses. When flow is less than infinite, and uptake is finite, then there is a gradation in concentration from arterial to venous end, a case of partial flow-limitation to uptake, so that the formula above overestimates the unidirectional uptake, because the local capillary concentration is less than C_A; with a 50% arterio-venous difference, the error is actually less than 50%, more like a third. The expression developed by Crone and Renkin accounts for the axial diminution in concentration, so that the arteriovenous extraction, E = (C_A − C_V)/C_A, can be correctly explained by the PS_Cap and the flow F whenever the extravascular volume is large enough that return flux is negligible:

$$\mathrm{E}=1-exp(-\mathrm{P}{\mathrm{S}}_{\text{Cap}}/\text{F})$$

Accounting for axial concentration gradients is even more important when there is return flux of tracer from the tissue to the blood, as there is in both the heart and the brain for the various glucoses. In the brain, the extractions of tracer d-glucose during single capillary passage are about 50% at low blood glucose levels and diminish toward 10% at levels 2 to 3 times normal, indicating that the escape from the capillary is large, and is carrier-mediated. In the heart, the instantaneous initial tracer extractions are 30% to 60% but are unaffected by the blood glucose level, since permeation is through aqueous channels between endothelial cells, and no carrier is involved. In both organs, values of PS_Cap are similar to those for flow and therefore the intracapillary axial gradients must be accounted for. The subsequent major needs are accounting for rapid reflux from tissue to blood, which the expressions above do not do, and accounting separately for uptake by the parenchymal cells of the organ. This requires models having the following minimal features: axial concentration-dependence, two barriers with bidirectional fluxes at each, and finally, intracellular reaction. Such models are now available.

The experimental requirement is that whenever there is partial flow limitation, flow should be measured independently of the glucose uptake. Receptor ligands and metabolites which have a PS_Cap of 0.5 ml g⁻¹min⁻¹ or greater are in this class.

The heterogeneity of regional flows in both brain and heart tends to about 30% (standard deviation/mean). There is also heterogeneity in partition coefficients and in volumes of distribution. The degree of error introduced by assuming homogeneity depends on the model, the particular parameter being evaluated, and the values of the parameter; errors in estimates of transport rates in the heart range from 30% to several hundred percent.

For accurate parameter estimation the exact shape of the input function must be known. To determine this shape, some method of measuring the arterial concentration is required. Of the three techniques used currently, only the direct arterial sampling is fully satisfactory for nongaseous tracers. The arterialized venous blood method, which employs hand-warming to open shunts, is valid only for steady-state concentration measurements. Sampling the left ventricular blood concentration using dynamic PET is satisfactory for those systems that can sample gated data at intervals less than 5 sec.

Once the input and residue curves are acquired with sufficiently rapid sampling, the validity of the fitting procedure and the appropriateness of the model come into question. Unfortunately, due to the presence of 15% to 25% noise, it is difficult to define parameters accurately with models with more than four rate parameters and three regions, unless more data than the single PET signal are used. Indeed, even the well-known Sokoloff model for brain glucose, or similar models used for evaluation of amino acid metabolism cannot, in practice, yield rate parameters with high accuracy, even though the numbers of parameters is small. One reason for this is the fact that the parameters of tissue blood volume and the forward (wash in) and reverse (wash out) transport rate constants are correlated, and the cross correlation terms should be included in an error analysis. Equally important, because the model is less complex than is essential for the conditions, the estimates of the rate constants are quite dependent on the form of the input; the error is least for the slowest rate constant, that of intracellular reaction. A priori information will help, but generally this is in the form of the fate of the label and not as additional tracer kinetic parameters or pool sizes. Of great benefit however are sets of data on regional flows (e.g., from ¹⁵O-water or “molecular microspheres,” indicators deposited in proportion to local flow), and on regional blood volumes.

In addition to flow and an accurate definition of the residue curve, some studies require measurement of the pool sizes. For example, when comparing the normal brain to the ischemic brain, the local blood volume might be a needed parameter that cannot be deduced accurately from kinetic analysis and should be measured if, for example, O₂ utilization and pH are to be inferred from tissue concentration measurements.

The ideal study with PET or single photon tomography requires an estimate of flow, pool sizes, and the fate of the label in addition to the actual measurement of the input function and the residue function. To neglect those parameters seriously compromises some studies (receptor concentration, pH), while in others one can use simplifying assumptions.

Areas of need for further investigation included:

1. Research the question of the physiologic variation of the capillary permeability-surface area product with flow, as it is a factor in determining the limits to adequacy of distributed models and lumped models.

2. Examine the importance of heterogeneity of flow, volume, and partition coefficients on the parameters derived from tracer detection methods. Both simulations and animal studies should be used.

3. Perform detailed examinations of the propagation of errors through model analyses.

4. Evaluate the sensitivity of the estimates of parameters to experimental protocols such as sampling rate and rate of bolus infusion.

5. Define the relations between kinetic parameters representing passive transport across barriers and carrier-mediated or active transport mechanisms.

V. New NMR techniques

A broad range of applications of multi nuclear magnetic resonance spectroscopy was presented by the formal speakers and discussants. Central questions concerned what kind of biomedical problems are solvable in principle and in current practice, what limitations and advantages does NMR have, how soon will these approaches be used in clinical medicine and will clinically useful information really be obtained. While the consensus was one of optimism, significant advances will require more integration of efforts amongst physicists, biochemists, physiologists, pathologists, clinicians, mathematicians and instrument technologists. While some semantic problems inhibit effective communication, the rapid phase of the learning curve is underway.

²³Na and ³⁹K NMR can be used to measure the asymmetry of their concentrations across the cell membranes. In contrast to covalently bonded NMR species, there is no natural chemically shifted form of either Na⁺ or K⁺ inside and outside the cell. Nontoxic paramagnetic compounds that do not penetrate the cell membrane (e.g., Dy(TTHA)³⁻) induce a chemical shift of Na⁺ and K⁺ in the extracellular fluid but not in the cytosol, thereby allowing the estimation of amounts outside versus inside.

Inside erythrocytes the NMR-visible Na⁺ appears to have the same concentration that is obtained by ordinary chemical analysis. However, in isolated perfused rat hearts, the intracellular Na⁺ estimated by NMR is 1 mM, while Na-sensitive microelectrodes indicate free Na⁺ concentrations of 6 to 10 mM, and atomic absorption spectometry gives total intracellular Na⁺ concentration (bound + free) of 20 to 30 mM. These results are quite provocative, being contrary to the idea that free ions should be visible by NMR. Even though quantitation may not be accurate, Na⁺ NMR may be quite useful; for example, an increase in myocardial Na⁺ can be detected after ouabain inhibition of the Na pump, a situation in which the total Na concentrations rise toward 100 mM.

A solution to an important problem limiting the application of spin transfer methods to the measurement of unidirectional reaction rate constants (and fluxes) was reviewed. These NMR methods provide unique information on the actual forward and reverse fluxes of steady-state metabolic reactions in intact cells and organs. The saturation of multiple species reduces more complex kinetic exchanges into simple ones, with a corresponding simplification of data collation and interpretation. The measurement of the flux between creatine phosphate and ATP, catalyzed by creatine kinase, is complicated by the possibility of exchanges between ATP and other species (for example, ATP hydrolysis and oxidative synthesis). If the latter occurs at a significant rate, then a two-site exchange model will not fit the data, as has been shown. However, saturation of the P_i effectively removes the ATP-P_i exchange from consideration, effectively reducing a three site exchange into a two site problem. In general, by appropriate saturation of N-2 exchange sites, a complex kinetic problem can be reduced to a two-site problem. Obviously these spin transfer methods must be applied only to steady-state systems (which is suitable for stable living systems on the time scale of these measurements). Caution must still be applied since interpretation of spin life times (which are what are actually measured in these experiments) as chemical ones is model dependent.

¹³C-NMR applications were briefly surveyed. Since ¹³C is present in natural compounds only as 1.1 % abundance, natural abundance ¹³C-NMR spectroscopy provides information only on highly concentrated compounds, e.g., fatty acids, triglycerides, and glycogen. It is possible to measure the degree of unsaturation of the fatty acid moiety in animals and in human limbs. Changes in the content of these stored metabolic fuels is another application that has been used only a little. Enrichment of specific metabolites (and even of specific carbons in metabolites) offers enormous opportunity for detailed metabolic investigations including the precursor-product relationships by time course studies. Additional information with greater sensitivity than the ¹³C observations can be obtained from ¹H NMR observations of specifically labeled ¹³C compounds taking advantage of information on the coupling of the protons attached to the ¹³C label with and without selective ¹³C irradiation. One of the problems is that the power of this approach has about 60-fold less sensitivity than the ¹H nucleus used for imaging. An area of future development is spin transfer kinetic measurements that merit considerable effort.

The approaches to obtaining chemical shift images, that is, spectroscopic information on a regional anatomic basis, range from the relatively simple use of surface coils, with and without depth resolving pulses, to some methods of chemical shift metabolic mapping. Examples were shown of human head and dog torso maps of ¹H of water vs fat. Well-resolved ³¹P spectra were obtained in 10 min from an estimated volume of 10 ml from a more or less planar region a few centimeters below the surface of a human brain using a surface coil technique. Application of proton chemical shift imaging for lactate and other metabolites, and of ¹⁹F images for the distribution of exogenous fluorocarbon compounds were also mentioned. The potential is very great; but the limiting factor is the current need for increased access to instruments with these capabilities.

The task of defining the range of biomedical and clinical problems that would have a reasonable probability of being addressed by these and other NMR methods remains. One example would be to use information on high-energy phosphate metabolites in the heart in defining operative risk and surgical strategy in patients with valvular disease. There are strong incentives to develop, test and apply the concept of metabolic stress tests, for these should give information currently not available by any other technique. A way to quantify hydroxyproline residues could be applied to monitor the progressive collagen deposition in myocardial fibrosis that eventually ends in heart failure. The metabolic basis of the fulminant failure characteristic of previously stable aortic stenosis could profitably be explored by a number of NMR techniques. Skeletal muscle can be readily studied in limbs, and so provide a sampling of intracellular metabolism in generalized metabolic disorders, such as shock and acidotic states.

For these and other applications of the technologies, it will be necessary to foster further communication among the various disciplines involved. Difficult problems remain, and advances will come mostly from cohesive interdisciplinary groups.

VI. Animal models for studies with NMR and PET

The choice of animal models is dictated by the specific question that is to be addressed and by the stage of development of experimental techniques and mathematical models to handle complex systems. The choice of test models for cardiac disease is governed by two requirements. The first is to obtain a quantitative estimation of flux of carbon or other atoms through specific pathways, particularly in discrete myocardial regions. The second, is to obtain high-quality images that will provide the clinician with an ongoing assessment reflective of the functional and/or metabolic status of the ventricular myocardium in clinical disease states.

To obtain quantitative estimates of metabolic pathway activity it is necessary to carry out studies in rather simple systems such as isolated perfused hearts. Such studies will allow verification of estimates made with noninvasive techniques by standard biochemical techniques, and based upon these studies, predictions of behavior in more complex systems can be made. NMR and PET studies have utilized isolated perfused hearts and in some cases standard biochemical and radionuclide analysis to verify the results.

More complex systems require such verification. This places restrictions on the animal model chosen for study. Desired features are those which allow direct measurement of variables, such as controllable or measurable coronary flow, mechanical function and arterio-venous differences in substrates and metabolites. Preferably these should be measurable within specific ventricular regions.

From perfused hearts one progresses often to more complex systems such as anesthetized open-chest swine, which have few coronary arterial collaterals, then to dogs with greater collateralization. The ultimate models require conscious animals in which regional function, regional coronary blood flow, and regional coronary venous blood samples can be obtained. These models have the advantages of allowing long-term studies under basal conditions.

Three conscious animal models meet the requirements: the dog, the baboon, and the horse. For some specific studies the dog appears not to be a suitable model. Too little is known about the horse as a model for human disease to evaluate it for use in a wide variety of studies. Ultimately, although expensive, nonhuman primates may be most useful for such studies, as they may provide the closest approximation to humans.

The heart disease most amenable to study and most studied with NMR and PET is myocardial ischemia. Problems that may be addressed include the effects of repeated ischemic insult vs long-term moderate ischemia; platelet thrombosis; increased myocardial oxygen demand in the presence of stenosis, drugs such as Ca⁺⁺ blockers, β-blockers, and vasodilators; effects of ischemia in hypertrophied or myopathic muscle; and the rate of coronary collateral vessel development and stimuli that may modify this rate.

Another area that may be assessed in a quantitative manner with imaging is myocardial hypertrophy. Major questions concern its genesis, its reversibility, and the association with myopathic features.

Qualitative estimates of myocardial metabolic activity appears to be a more easily obtainable goal. There are a number of clinically important problems that may be addressable with NMR and PET. These include assessment of graft patency and the effects of thrombolysis and angioplasty; determining time of origin of myocardial ischemia, the rate of its progress, and the reversibility of damage; accurate estimations of regional coronary flow; assessment of efficacy of therapeutic interventions; “ischemia at a distance,” and metabolism in noninfarcted tissue; and the magnitude of pump failure. For managing surgical patients it would be advantageous to develop imaging to assess postsurgical mechanical function, coronary flow, and metabolism in order that recovery can be monitored, indications for weaning from cardiopulmonary bypass be determined, and the efficacy of pharmacologic agents assessed.

Myocardial metabolic activity has been estimated with PET with some success. Whereas deoxyglucose may reflect glucose transport, fluoroacetate should be studied as a possible marker for the redox status of cells. In studies of lipid metabolism, appropriate mathematical treatments of flux kinetics must be developed. ¹¹C-palmitate has been the most widely used for fatty acid metabolism, but major problems remain for estimating the quantities of label that enter and leave cells, are oxidized and are stored as triglycerides. Provided that flow-related extraction can be adequately accounted for, β-methyl fatty acids may provide an estimation of β-oxidation and hence mitochondrial NAD⁺/NADH ratios. Studies using short chain fatty acids may be particularly useful in bypassing mitochondrial activation and transesterification pathways. Studies of intracellular distribution of label would be aided by use of specific inhibitors to block intermediate and branch points in fatty acid metabolism. By working out the kinetics of lipid metabolism in this way, PET techniques then could be applied to other abnormalities involving fatty acid metabolism such as carnitine myopathies and patients with Reyes disease.

The likelihood that ¹³N can become a useful tool for quantitative estimates of myocardial metabolism is less clear than is the case of ¹¹C and ¹⁸F compounds. ¹³NH₃ as a blood flow marker is seen as having advantages and problems because it is metabolized. Use of ¹³N-labeled amino acids holds some promise for studying protein synthesis but because the amino acids are recycled and protein turnover is generally slow, their usefulness is limited. Further complicating such studies is the likelihood that amino acids are incorporated into rapidly turning over proteins in myocardial cells other than myocytes and that leucine oxidation is NAD⁺ dependent, and inhibited by long chain acylcarnitines.

The major limitation of PET for study of energy metabolism is the inability to monitor high-energy phosphate concentrations. Thus, joint NMR and PET studies may be of particular use as they would allow simultaneous or sequential estimates of substrate flux, ATP, CP, creatine, and pH.

The virtues of NMR in myocardial metabolic studies relate to the capability to detect ATP and creatine phosphate and to provide estimates of cellular pH, phosphate flux, and coupling between ATP and creatine phosphate. Major limitations are the low spatial resolution of imaging anything other than protons; the poor time resolution; the need to develop imaging capabilities for other elements, such as Na; and to improve surface coil techniques to give local information for a wider variety of elements including ³¹P. Presently, the tradeoff is between large magnets with less sensitivity vs smaller probes with greater spatial resolution.

VII. Needs and opportunities

New imaging techniques offer new insight into metabolism

The utility of new methods for measuring the transport and transformation of substrates of metabolism in vivo should be seen in a broad context. Test tube experiments, studies on tissue slices, cultured cells, and isolated organs, provide information from cells, organelles, or enzymes in nonphysiologic conditions. Tracer experiments requiring tissue sampling at chosen times for estimating content by sample counting or quantitative autoradiography provide only one time point from each analysis rather than the whole time course. In contrast NMR can provide data encompassing changes of physiologic state. PET can provide the whole sequential time course of the content of a tracer in many small regions of an organ following a single injection. These capabilities offer the first real opportunities for studying metabolism in vivo, rather than in vitro, via localized information in intact integrated tissues and organs of humans and animals.

An image must be thoroughly understood to be useful

There is, particularly in NMR, a need to determine the physical basis of the observed signals. The signals detected by noninvasive methods of emission tomography and NMR arise from physical and chemical processes whose identity can be confused due to the fact that the same signal differences (e.g., image contrast) can arise from entirely different mechanisms, e.g., a change in NMR signal due to flow rather than a change in intrinsic tissue properties. An accumulation of an amino acid analog shown by PET might be due to a nonspecific change in membrane permeability rather than to a change in active transport. Therefore research focused on clarifying the physical and kinetic bases of signals detected by both NMR and PET is needed. Because of the complexities of these physical processes, how they change in pathologic states must be investigated as well.

The clinical utility of metabolic information needs to be affirmed

Recent technological developments such as NMR and PET potentially allow the in vivo study of myocardial metabolism. Research is needed to examine the value of this newly available information for distinguishing between normal and abnormal states and responses. For example, will this information aid in better understanding of commonly measured expression of cardiac disease? Will it be useful for determining more accurately the severity of a disorder, or for distinguishing between reversible and irreversible disease states? Can the effects of therapeutic interventions be monitored and their mechanisms elucidated with these new modalities? Do PET and NMR provide information unobtainable with other techniques? (For example, can silent cardiac disease be detected?) Can they contribute to understanding the pathophysiology and pathogenesis of disease?

Modeling analysis is essential, but the field needs much development. Particular needs are as follows

Tests of the appropriateness of models

A simple model is highly advantageous from the points of view of ease of comprehension and speed of analysis, but runs the risk of providing inaccurate or invalid interpretations. One strategy is to evaluate any simplified model carefully in terms of a relatively detailed, comprehensive model which, so far as is possible, correctly accounts for all of the known features of the system.

The development of large, integrated models

Each requires major research efforts as well as major technical efforts. The benefit is that a comprehensive model provides a descriptive and theoretical base for many experiments; the development process itself leads to identification of components of the system about which more needs to be known.

Mechanisms for model dissemination to prospective users

Models of important metabolic and transport systems should be made available to the research community in order to extend each investigator's capabilities to consider complicated systems and situations. This should allow gains in efficiency and selectivity in the choosing and designing of experimental studies.

Use models for the quantitative analysis of data via parameter identification, as well as for the design of experiments

Each model should attempt to encompass both the normal and the pathophysiologic state. Without this, a clinical user would have to choose a model based on his diagnosis, rather than have the modeling assist him in making it. Such a severe demand necessitates using complicated models, and that it would be risky to use different reduced forms in different states. The models should be developed to the point where they are useful for clarifying metabolic mechanisms in abnormal states.

Biochemical regulation is an important research target

Data analyses using metabolic tracers are difficult or inexact because there is poor information on the prevailing metabolic state of the tissue studied and the sensitivity of the cells to hormones in normal and pathologic states. Therefore, it is most important to define the regulatory processes, in both physiologic and pathophysiologic contexts. Until now, knowledge on metabolic regulation has mainly been extrapolated from in vitro observations to the in vivo situation. The future implications for metabolic imaging will help: (1) to define metabolic regulation under true physiologic conditions; (2) to assign quantitative importance to a variety of redundant regulatory mechanisms; (3) to uncover new mechanisms of control; (4) to quantitate the physiologic and pathologic responses to biochemical change; and (5) to examine the regulatory roles of membrane receptors and transporters, and of membrane function in general on biochemical regulation.

Remote or nonlocal influences on the metabolic state require elucidation in a wide variety of states

Nutrition, exercise, and training can affect myocardial and peripheral metabolic and physiologic responses to stress. The mechanisms for these and related responses may be studied by nontraumatic approaches such as PET and NMR, applied for example to skeletal muscle or to brain to determine changes in perfusion, energetics, and function. From such insights, the regulatory and adaptive processes would be better understood and the clinical importance of such changes defined. Conversely, cardiac disease can alter the peripheral metabolic and physiologic state. Evaluation of the peripheral metabolic and physiologic state could therefore be useful for dinstinguishing between adequate and inadequate adaptations (compensation or decompensation) to cardiac disease and, hence, allow assessment of severity of cardiac disease (e.g., congestive heart failure).

An observable metabolic status should be explicable in terms of gene activity

The importance of cardiac responsivity to its supply of metabolic fuels may reflect an overall adaptation of the heart, with respect to: (1) substrate preference, (2) efficiency of energy production, and (3) energy utilization by the contractile proteins. At any particular time, the integrative behavior or metabolic expression of the heart represents a long-term consequence of gene activity or nutritional changes, as opposed to short-term variations in the substrate supply and acute responses via normal pathway regulatory mechanisms. Therefore there is a clear need to understand the influences of nutritional components, under both acute and chronic conditions, on the mechanisms of gene expression, i.e., induction and repression of gene products, as related to metabolic signals in vivo, especially those that lead to alterations in contractile velocity and efficiency. Long-term noninvasive studies should be useful in evaluating such changes.

Use three-dimensional integrators of information

In view of the above recommendation regarding regional metabolic information the development of techniques to provide three-dimensional information from both PET and NMR should be encouraged. These should include taking into account information from all available planes or fields in reconstructing images in order to obtain higher spatial resolution, which is especially important for PET and for non-proton NMR images.

Chemical and tracer compounds must be developed and evaluated

Encourage the development, testing, and validation of radiopharmaceuticals and the techniques for evaluating the characteristics of radiopharmaceuticals or metabolic tracers. Likewise, encourage the development of contrast enhancing agents, shift reagents, receptor ligands, etc. An example is the use of shift reagents to distinguish intracellular from extracellular sodium. Each agent usually requires years of effort in chemistry validation and in kinetic analysis before it can be considered as a known compound for routine use.

Support the training of investigators in this expanding field

The relevance of basic sciences to the study of medicine has a clear impact in the field on noninvasive imaging. Training physician-scientists in biochemistry, physiology, and kinetics must begin in the early years of medical education by acquainting students with these important new clinical approaches. Both PET and NMR are unique tools to enhance awareness of the importance of biochemistry and physiology at the organ level. The rewards of disseminating this information to medical and graduate students lie in future approaches to problem solving and full development of the potential of these powerful techniques. Postdoctoral training efforts in collaborative research amongst physicians, spectroscopists, biochemists, and physiologists should be strongly encouraged.

Plan studies using both PET and NMR

Because of the complementarity of the information provided by PET and NMR studies, experimental protocols applying both techniques to the same system should be given strong encouragement. The prerequisite for this is that both techniques should be thoroughly evaluated with respect to the quantitative interpretation of the observed signals, but it is recognized that attempting to obtain the same information by both PET and NMR will provide a check on the adequacy of both techniques. In many other situations, the data from each will be of a different character but will be integrated in the assessment of the biochemical or physiologic states.

Appendix A: Agenda

Chairman: James Bassingthwaighte
Monday, September 17
Welcome	Barbara Packard
	Director, DHVD, NHLBI
Introduction	James Bassingthwaighte
Session I: Metabolism overview: current dilemmas and future directions
Moderator: Jeanie McMillin-Wood
Carbohydrate interconversions and energy production	Heinrich Taegtmeyer
Fatty acid, triacylglycerol, and ketone body metabolism	Loran Bieber
Nitrogen metabolism and protein synthesis/degradation mechanisms	Radovan Zak
Metabolic pathways of energy metabolism: assessment by nuclear magnetic resonance	Truman Brown
Metabolic tracer imaging with positron-emission tomography	Burton Sobel
Session II: Kinetics of substrate transport and reaction
Moderators: James Bassingthwaighte and Kenneth Zierler
Overview: Processes of delivery, transport, reaction, and retention	James Bassingthwaighte
The modeling process	Kenneth Larson
Heterogeneity within the observed regions: physiological basis and effect on estimation of rates	Ludvik Bass
The acquisition of data for quantitative analysis	Thomas Budinger
Model simplification: complexity versus reduction	Michael Graham
Panel and open discussion: recommendations for developments in analysis of metabolic events via images
Critique of session	Kenneth Zierler
Discussion	Ludvik Bass, James Basssingthwaite, Thomas Budinger, Albert Gjedde, Michael Graham, and Kenneth Larson
Monday evening, September 17 (concurrent sessions)
Session III. Influences of flows and transmembrane transport on tracer uptake retention
Moderators: Thomas Budinger and Albert Gjedde
Influences of flow on substrate uptake	James Holden
Error in parameter estimates with variations in flow	Sung-Cheng Huang
A proposed model system for studying receptor kinetics	Kenneth Krohn
Estimation of membrane transport rates in the absence of regional flow measurement	Carl Goresky
Effects of reflux from the cell on the interpretation of deposition images	Michael Phelps
Moderators' summary: Thomas Budinger and Albert Gjedde
Session IV. NMR kinetic measurements
Moderators: Truman Brown and Joanne Ingwall
Use of shift reagents for kinetics of ion transfer	Charles Springer
Magnetization transfer techniques	Kamil Ugurbil
Metabolic studies using 13C	Jan den Hollander
Spectroscopic localization techniques	Felix Wehrli
Moderators' summary: Truman Brown and Joanne Ingwall
Tuesday, September 18
Session V. Myocardial disease
Moderator: Heinrich Schelbert
Critical problems in myocardial disorders
Need for noninvasive studies of regional myocardial metabolism in clinical cardiology	Edmund Sonnenblick
Biochemical abnormalities in cardiac diseases	Arnold Katz
Approaches to future studies using NMR
Spectroscopy in vivo	Britton Chance
NMR imaging: potential for metabolic studies	Gerald Pohost
Approaches to future studies of the disease process using positron emission tomography
Assessment of substrate metabolism	Heinrich Schelbert
Krebs cycle, oxygenation, and oxidation	Terry Jones
Assessment of cardiac receptors	William Eckelman
Synopsis and summary	Henry Wagner
Session VI. Animal model studies
Moderator: Truman Brown
What is needed in an animal model	Michael Rovetto
Specific cardiac models	James Scheuer
Metabolic and physiological model studies using PET	Jeanie McMillin-Wood
Metabolic and physiological model studies using NMR	Robert Balaban
Future opportunities	Thomas Smith
Summary: meeting overview	Philip Randle

Appendix B: Participants

Robert S. Balaban, Ph.D., Senior Staff Fellow, National Heart, Lung, and Blood Institute, Room 6N307, Building 10, Bethesda, MD 20205

Ludvik Bass, Ph.D., Professor of Mathematics, University of Queensland, Brisbane, Australia Q 4067

James B. Bassingthwaighte, M.D., Ph.D., Professor, University of Washington WD-12, Seattle, WA 98195

Myrwood C. Besozzi, M.D., Director of Nuclear Cardiology and Cardiac NMR Imaging, University of Wisconsin, Cardiology Department, 600 Highland Avenue, Madison, WI 53792

Loran L. Bieber, Ph.D., Professor of Biochemistry, Department of Biochemistry, Michigan State University, East Lansing, MI 48824

Truman R. Brown, Ph.D., Director of NMR and Medical Spectroscopy, Fox Chase Cancer Center, 7701 Burholme Avenue, Philadelphia, PA 19111

Thomas F. Budinger, M.D., Ph.D., Henry Miller Professor of Research Medicine, Donner Laboratory University of California, University of California at Berkeley, Donner Laboratory, Rm. 230, Berkeley, CA 94720

Britton Chance, Ph.D., Director, Institute for Structural and Functional Studies, D 501 Richards Building/G4, Philadelphia, PA 19104

Sheila M. Cohen, Ph.D., Research Fellow, Department of Biophysics, Merck Institute for Therapeutic Research, P.O. Box 2000, R80C 25N, Rahway, NJ 07065

William C. Eckelman, Ph.D., Chief, Radiopharmaceutical Chemistry Section, Department of Nuclear Medicine, National Institutes of Health, Building 10, 1C-488, Bethesda, MD 20205

S. John Gatley, Ph.D., Associate Professor of Medical Physics, Department of Medical Physics, 1530 Medical Science Center, University of Wisconsin, 1300 University Avenue, Madison, WI 53706

Raymond Gibbons, M.D., Consultant in Cardiovascular Diseases and Internal Medicine, Divsion of Cardiovascular Diseases, W16B, Mayo Clinic, Rochester, MN 55901

H. F. Gilbert, Ph.D., Associate Professor of Biochemistry, Baylor College of Medicine, Houston, TX 77030

Albert Gjedde, M.D., Ph.D., Associate Professor of Physiology, Medical Physiology Department A, Copenhagen University, The Panum Institute, Blegdamsvej 3, Copenhagen, Denmark 2200

Carl A Goresky, M.D., Professor of Medicine, McGill University, Montreal General Hospital, 1650 Cedar Avenue, Montreal, Quebec H3G 1A4

K. Lance Gould, M.D., Professor and Director, Division of Cardiology, University of Texas, HSC at Houston, 6431 Fannin, 1.236 MSMB, Houston, TX 77030

Michael M. Graham, Ph.D., M.D., Assistant Professor of Medicine and Radiology, Nuclear Medicine RC-70, University of Washington, Seattle, WA 98195

Robert J. Herfkens, M.D., Assistant Professor of Radiology, Department of Radiology, Duke University, Box 3808, Durham, NC 27710

Julien I. Hoffman, M.D., Professor of Pediatrics & Physiology, University of California, San Francisco, 1403 HSE, San Francisco, CA 94143

James E. Holden, Ph.D., Associate Professor of Medical Physics and Radiology, Room 1530 Medical Sciences Center, University of Wisconsin, Madison, 1300 University Avenue, Madison, WI 53706

Jan den Hollander, Ph.D., Staff Scientist, Phillips Medical Systems, Eindhoven, Netherlands 560MD

Sung-Cheng Huang, D.Sc., Associate Professor of Biophysics, University of California at Los Angeles, School of Medicine, Los Angeles, CA 91607

Joanne S. Ingwall, Ph.D., Associate Professor of Physiology & Biophysics, Harvard Medical School, Brigham and Women's Hospital, 75 Francis Street, Boston, MA 02115

Terry Jones, M.Sc., Head, Physics Isotope Section, Hammersmith Hospital, Ducane Road, London W12, UK

Arnold M. Katz, M.D., Professor of Medicine, Head, Cardiology Division, University of Connecticut Health Center, Farmington, CT 06032

Kenneth A. Krohn, Ph.D., Associate Professor of Radiology, Division of Nuclear Medicine RC-70, University of Washington, Seattle, WA 98195

Martin J. Kushmerick, M.D., Ph.D., Associate Professor/Physiology/Biophysics, Head of Metabolite Research in the Division of NMR, Harvard Medical School, 25 Shattuck Street C2-547, Boston, MA 02115

Ken Larson, Ph.D., Biomedical Computer Laboratory, Washington University, 700 S. Euclid Avenue, St. Louis, MO 63110

Steven M. Larson, M.D., Chief Nuclear Medicine, National Institute of Health, ACRF, Building 10, Room 1C401, Bethesda, MD 20205

Giovanni Lucignani, M.D., Visiting Fellow, Laboratory of Cerebral Metabolism, National Institute of Mental Health, Building 36, Room 1A05, Bethesda, MD 20205

Craig Malloy, M.D., Assistant Professor of Medicine and Cardiology, Department of Medicine, University of Texas Health Science Center at Dallas, 5323 Harry Hines Boulevard, Dallas, TX 75235

Jeanie B. McMillin-Wood, Ph.D., Associate Professor of Medicine and Biochemistry, Baylor College of Medicine, Texas Medical Center, Department of Medicine, Houston, TX 77030

John Mowbray, Ph.D., Lecturer in Biochemistry, Department of Biochemistry, University College London, Gower Street, London, UK WC1E 6BT

Nizar A. Mullani, B.S., Assistant Professor of Medicine, University of Texas Medical School, Div. of Cardiology, 6431 Fannin, MSMB 1.246, Houston, TX 77025

Ray L. Nunnally, Ph.D., Assistant Professor of Radiology, Director of Biomedical Magnetic Resonance, University of Texas Health Science Center at Dallas, 5323 Harry Hines Boulevard, Dallas, TX 75235

Brian D. Pate, Ph.D., Director, Program of Positron Emission Tomography, University of British Columbia, 2211 Wesbrook Mall, Vancouver, B.C., Canada V6T 2B5

Michael E. Phelps, Ph.D., Division of Biophysics, University of California at Los Angeles, School of Medicine, Los Angeles, CA 90024

Gerald M. Pohost, M.D., Professor of Medicine and Radiology, Department of Medicine, The University of Alabama in Birmingham, University Station, Birmingham, AL 35294

Marcus E. Raichle, M.D., Professor, Neurology and Radiology Division, Washington University, St. Louis, MO 63110

Philip J. Randle, M.D., Ph.D., Professor, Department of Clinical Biochemistry, University of Oxford, John Radcliff Hospital, 0X3 9DU, UK

Russell Reeves, M.D., Assistant Professor of Medicine, University of Alabama, Birmingham, Room 338, LHR, Birmingham, AL 35294

Michael J. Rovetto, Ph.D., Associate Professor, Department of Physiology, University of Missouri, Columbia, School of Medicine, M411 Medical Sciences, Columbia, MO 65212

Heinrich R. Schelbert, M.D., Professor of Radiological Sciences, Division of Nuclear Medicine, University of California, Los Angeles, School of Medicine, Los Angeles, CA 90024

James Scheuer, M.D., Professor of Medicine and Physiology, Albert Einstein College of Medicine, Montefiore Medical Center, 111 E. 210th Street, Bronx, NY 10467

Andrew Selwyn, M.D., FACC, Director Cardiovascular Division, Brigham and Women's Hospital, Harvard Medical School, 75 Francis Street, Boston, MA 02115

Thomas W. Smith, M.D., Professor of Medicine, Cardiovascular Division, Brigham and Women's Hospital, Harvard Medical School, 75 Francis Street, Boston, MA 02115

Burton E. Sobel, M.D., Director, Cardiovascular Division, Washington, University, School of Medicine, 660 S. Euclid, Box 8086, St. Louis, MO 63110

Louis Sokoloff, M.D., Chief, Laboratory of Cerebral Metabolism, National Institutes of Mental Health, Building 36, Room 1A-07, Bethesda, MD 20205

Edmund H. Sonnenblick, M.D., Professor, Department of Medicine, Albert Einstein College of Medicine, 1300 Morris Park Avenue, Bronx, NY 10461

Charles S. Springer, Jr., Ph.D., Associate Professor, Department of Chemistry, State University of New York at Stony Brook, Stony Brook, NY 11794

Andre Syrota, M.D., Ph.D., Professor of Biophysics and Nuclear Medicine, Service Hospitalier Frederic Joliot, Orsay 91406, France

Heinrich Taegtmeyer, M.D., D.Phil., Assistant Professor of Medicine, Division of Cardiology, University of Texas Medical School at Houston, 6431 Fannin, Houston, TX 77030

Kâmil Uǧurbil, Ph.D., Associate Professor, Department of Microbiology, Gray Freshwater Biological Institute, University of Minnesota, Navarre, MN 55392

Richard L. Veech, M.D., D.Phil., Chief, Lab of Metabolism, National Institute Alcohol Abuse and Alcoholism, 12501 Washington Avenue, Rockville, MD 20852

Henry N. Wagner, Jr., M.D., Division of Nuclear Medicine, Johns Hopkins Medical Institution, Room 1102, 615 North Wolfe Street, R 2001, Baltimore, MD 21205

Felix W. Wehrli, Ph.D., Manager MR Applications, General Electric Company, Medical Systems Group, P.O. Box 414, Milwaukee, WI 53201

Radovan H. Zak, Ph.D., Professor Medicine and Pharmacological and Physiological Sciences, University of Chicago, 950 East 59th Street, Chicago, IL 60637

Kenneth L. Zierler, M.D., Professor of Physiology and Medicine, The Johns Hopkins University, School of Medicine, 725 N. Wolfe Street, Baltimore, MD 21205, NHLBI Representatives:

NHLBI Representatives:

Barbara L. Packard, M.D., Ph.D., Director, Division of Heart and Vascular Diseases, National Heart, Lung, and Blood Institute, Federal Building, Room 416A, Bethesda, MD 20205

Eugene R. Passamani, M.D., Associate Director for Cardiology, Division of Heart and Vascular Diseases, National Heart, Lung, and Blood Institute, Federal Building, Room 3C-02, Bethesda, MD 20205

Thomas Robertson, M.D., Chief, Cardiac Diseases Branch, Division of Heart and Vascular Diseases, National Heart, Lung, and Blood Institute, Federal Building, Room 3C-02, Bethesda, MD 20205

Zena McCallum, Scientific Project Officer, Cardiac Diseases Branch, Division of Heart and Vascular Diseases, National Heart, Lung, and Blood Institute, Federal Building, Room 3C-06, Bethesda, MD 20205