former is still a very potent drug; howev- er, because of a scarcity of sample, di- demnin A has not yet been evaluated at optimum dosage levels for its antitumor effect in vivo. Didemnin C, the trace component, has not been available in quantities adequate for extensive testing. Interest in the chemotherapeutic po- tential of the didemnins is heightened by recent investigations (9) resulting in the structure elucidation of didemnins A, B, and C (Fig. 1). Novel aspects are a new structural unit for depsipeptides, hy- droxyisovalerylpropionate, and a new stereoisomer of the highly unusual amino acid statine. It is particularly noteworthy that didemnins B and C are simple deriv- atives of didemnin A. That the biological activities of the didemnins can be dra- matically altered by slight chemical changes bodes well for the development of a useful therapeutic agent. KENNETH L. RINEHART, JR. JAMES B. GLOER Roger Adams Laboratory, University of Illinois, Urbana 61801 ROBERT G. HUGHES, JR. Department of Cell and Tumor Biology, Roswell Park Memorial Institute, Buffalo, New York 14263 HAROLD E. RENIS J. PATRICK MCGOVREN EVERETT B. SWYNENBERG DALE A. STRINGFELLOW SANDRA L. KUENTZEL LI H. LI Upjohn Company, Kalamazoo, Michigan 49001 References and Notes 1. Presented in part,at the 3rd International Sym- posium on Marine Natural Products (Interna- tional Union of Pure and Applied Chemistry and Societe Chimique de Belgiques), Brussels, 16 September 1980. 2. P. D. Shaw, W. 0. McClure" G. Van Blaricom, J. Sims, W. Fenical, J. Rude, in Food-Drugs from the Sea Proceedings 1974, H. H. Webber and G. D. Ruggieri, Eds. (Marine Technology Society, Washington, D.C., 1976), pp. 429-433; K. L. Rinehart, Jr., R. D. Johnson, I. C. Paul, J. A. McMillan, J. F. Siuda, G. E. Krejcarek, ibid., pp. 434-442. 3. K. L. Rinehart, Jr., et al., Pure Appl. Chem. 53, 795 (1981). 4. M. T. Cheng and K. L. Rinehart, Jr., J. Am. Chem. Soc. 100, 7409 (1978). 5. G. T. Carter and K. L. Rinehart, Jr., ibid., p. 7441. 6. By Dr. Francoise Lafargue, Laboratoire Arago, Banyuls-sur-mer, France, and Dr. Charles C. Lambert, Department of Biological Science, California State University, Fullerton. 7. H. E. Renis, Antimicrob. Ag. Chemother. 11, 701 (1977); ibid. 13, 613 (1978). 8. , unpublished data. 9. K. L. Rinehart, Jr., J. B. Gloer, J. C. Cook, Jr., S. A. Mizsak, and T. Scahill, J. Am. Chem. Soc. 103, 1857 (1981). 10. Supported in part by grants from the National Institute of Allergy and Infectious Diseases (AI 04769), the National Institute of General Medi- cal Sciences (GM 27029), and from the National Science Foundation (PCM 77-12584). We thank the governments of Colombia, Honduras, Mexi- co, Belize, and Panama for permission to carry out scientific research in their territorial waters. 22 September 1980; revised 6 February 1981 SCIENCE, VOL. 212, 22 MAY 1981 Gated Sodium-23 Nuclear Magnetic Resonance Images of an Isolated Perfused Working Rat Heart Abstract. Sodium-23 nuclear magnetic resonance images of phantoms and gated images of isolated perfused working rat hearts were obtained. By synchronizing the nuclear magnetic resonance process to the heartbeat, images were obtained at systole and at diastole. Since its inception (1), nuclear mag- netic resonance (NMR) imaging has pro- gressed from a curiosity to the point where it promises to become one of the more important diagnostic tools in medi- cine. Several examples of high-resolu- tion (millimeter) NMR images of pro- truding appendages of the human body, notably limbs and heads, have been pub- lished (2-4). Technological progress is being made, and NMR images of cross sections of the human torso may soon have a resolution comparable to that now obtained for heads and limbs. It is important that methods for using NMR imaging as a noninvasive diagnostic mo- dality in cardiovascular research be de- veloped. Gating the acquisition of NMR signals to the heartbeat (5) was essential in our experiment in order to overcome the problems posed by heart motion. In ad- dition to developing NMR technology suitable for imaging a beating heart, the physiological basis for providing con- trast between the heart and the blood must be identified. Proton NMR images based on proton density alone may pro- vide little contrast between blood and surrounding tissues. However, there is a significant difference between the con- centration of sodium in blood and that in healthy tissue. We sought, therefore, to produce 23Na NMR images of the heart. They would be negative images of the myocardium inasmuch as healthy tissue has a low sodium content compared to blood. For these experiments, we modified a Nicolet wide-bore NT 360 spectrometer (8.45 T) so that it would perform as an imaging instrument. This involved the addition of three computer-controlled digital-to-analog converters to vary the currents of the first-order gradient shim coils, and the composition of several routines to provide these controls during acquisition of NMR data and to allow reconstruction and display of images. The imaging method used was basically the projection reconstruction method (1) with the image plane defined by an adap- tation of the z-gradient oscillation meth- od. The thickness of the slice was estab- lished experimentally by use of a phan- tom consisting of a flat-bottom NMR tube with a thin layer (1 mm) of 100 mM NaCl on the bottom. By moving this phantom in the probe with the oscillating z gradient on, the thickness of the im- aged slice was found to be about 1.5 mm. The acquisition of each free-induction decay was triggered from the aortic pres- sure wave. A delay between the trigger from the pressure wave and acquisition is programmed in order to choose the instant within the cardiac cycle at which the acquisition is triggered. As has traditionally been the case, the image-producing system was validated by making 23Na images of phantoms. Figure la shows a diagram of the cross section of a phantom that consisted of a 20-mm outer diameter NMR tube filled with distilled water, into which were placed 5- and 2-mm outer diameter tubes containing 145 mM NaCl. Figure lb is the 23Na NMR image of this phantom (obtained at 95.25 MHz) resulting from 12 projections in the x-y plane and recon- structed by the standard back-projection method. Each projection required aver- aging of 320 free-induction decays. The images shown in this report are defined by a matrix of 64 by 64 pixels and were photographed from the screen of a Hew- lett-Packard 1304A, producing (unfortu- nately) minimal levels of contrast. We obtained images of an isolated perfused working rat heart, using the perfusion apparatus previously de- scribed (5). The only modification was that the suction cannula was raised so that the level of perfusate in the NMR tube was above the heart. Since the perfusate (modified Krebs-Henseleit bi- carbonate, pH 7.24) has a high sodium content, its presence external to the heart provided contrast to the ventricu- lar wall. Gated planar images were ob- tained of a midventricular slice of the isolated working heart of a 350-g Sprague-Dawley rat (Fig. 2). These im- ages were reconstructed from 12 projec- tions, each projection obtained by aver- aging 320 free-induction decays. Since the spin-lattice relaxation time of sodium is short (approximately 40 msec) in tis- sue, acquisitions can be closely spaced. One free-induction decay can be collect- ed during each heartbeat (approximately 230 msec). Thus each image required 0036-8075/81/0522-0935$00.50/0 Copyright © 1981 AAAS 935