Proc. Nati. Acad. Sci. USA Vol 76, No. 3, pp. 1140-1144, March 1979 Biochemistry Cross-sectional views of hemoglobin S fibers by electron microscopy and computer modeling (sickle cell hemoglobin fibers/thin sections/tannic acid embedding) ROBIN L. GARRELL, RICHARD H. CREPEAU, AND STUART J. EDELSTEIN Section of Biochemistry, Molecular and Cell Biology, Cornell University, Ithaca, New York 14853 Communicated by Helen M. Ranney, December 13,1978 ABSTRACT Fibers of deoxyhb S have been investigated by thin-section electron microscopy, utilizing a tannic acid embedding procedure. On the basis of numerous measurements of cross-sectional center-to-center distances for adjacent fibers in pairs or arrays, fiber diameters (mean + SD) of 205 ± 5 A in embedded cells and 212 ± 8 A in embedded hemolysates were obtained. This is in agreement with values obtained by conventional embedding procedures [Crepeau, R. H., Dykes, G., Garrell, R. L. & Edelstein, S. J. (1978) Nature (London) 274, 616-617]. The use of tannic acid has resulted in improved resolution of fiber cross sections, revealing individual strands of Hb S molecules. Because the section thickness corresponds to approximately one-fifth of the fiber helical repeat distance, the strands in projection superimpose to form characteristic image patterns. Additional superposition patterns arise in sections taken at small deviations from perpendicularity to the longitudinal fiber axis. These patterns are consistent with the 14-strand structure for hemoglobin S fibers [Dykes, G., Crepeau, R. H. & Edelstein, S. J. (1978) Nature (London) 272, 506-5101, as indicated by computer models of cross-sectional patterns for various thicknesses and angular deviations of sections. Studies on the structure of the fibers of Hb S have focused on two forms: a t170-a-diameter form with a 6-strand, stackeddisk structure first described by Finch et al. (1) and more recently investigated by Ohtsuki et al. (2); and a z200-a-diameter form with a helical structure recently described by Dykes et al. (3) in terms of a 14-strand model. A major question has been Which of these forms predominates in sickled cells? Because structural studies involving negatively stained preparations examine only a small percentage of the fibers present in a population, it may be difficult to infer what is occurring in intact cells from such studies. Nevertheless, the 14-strand fiber would appear to be the major form on the basis of its predominance in numerous fiber preparations from lysed sickled cells, gelled hemolysates, and bundles formed from stirred hemolysates (4). A more direct approach to the question would be to examine embedded preparations and, on the basis of apparent fiber diameters, attempt to discriminate between the alternative possibilities. Such an examination has been completed and indicates agreement, within the statistical limits of the measurements, between the mean (±SD) diameter (measured as center-to-center distance) for adjacent fibers in pairs and arrays of embedded and sectioned sickled cells, 217 i 11 A (a similar value was obtained for fibers in stirred hemolysates), and that of the 14-strand form of the fibers observed in negatively stained preparations, 208 + 10 A (5). Reported here are values in this range that have been obtained for stirred hemolysates and sectioned sickled cells embedded by new techniques. These results substantially strengthen the argument in favor of the 14-strand form as the major class of fibers in vivo, The publication costs of this article were defrayed in part by page charge payment. This article must therefore be hereby marked "advertisement" in accordance with 18 U. S. C. 1734 solely to indicate this fact. but the possibility that distortions might arise during the embedding and sectioning of sickled cells cannot be completely excluded. Perhaps the most direct evidence that could be obtained on the issue of the fiber form present in sickled cells would be to count the strands of Hb S molecules present in cross sections of embedded fibers. In extensive studies using conventional embedding techniques to determine the diameter of fibers in sickled cells and cell-free preparations, however, the fiber cross sections appeared as amorphous, continuous, near-circular or elliptical areas with no evidence of substructure (5). In an effort to improve the resolution in cross sections of the fibers, we have now carried out studies using the embedding procedure of Mizuhira and Futaesaku (6) involving tannic acid. With this procedure, individual strands of tubulin molecules have been revealed in microtubules and therefore reasonable expectations existed that similar resolution would be obtained with Hb S fibers. As reported here, the addition of tannic acid greatly enhances the detail visible in cross sections of Hb S fibers, both in cells and in hemolysates, although the most dramatic effects have so far been limited to the latter. Although detail is enhanced with tannic acid, it is still not possible to enumerate the strands directly as can be done for microtubules. The reason for this difference is that, with microtubules, the individual strands or protofilaments are strictly parallel to the microtubule axis; therefore, the contributions at various depths in cross sections reinforce to give clearly defined strands. In the case of Hb S fibers, the strands follow a helical course with a repeat distance of about 3000 A. Thin sections (approximately 600 A in depth) will contain individual strands traversing approximately onefifth of a helical repeat from the top to the bottom of the section. Thus, the contributions of one strand at the top surface of a section will superimpose with the contributions of a different strand from the bottom surface of the section, and various superposition patterns will result. By generating computer models it has been possible to show that the distinctive fiber cross-sectional patterns observed in thin sections closely correspond to the results expected for the 14-strand fibers. MATERIALS AND METHODS Sample Preparation. Whole blood was obtained from patients with sickle cell anemia. After three washes with 1% NaCl, unlysed cells were conserved for whole-cell preparation. Solution of cells were deoxygenated under N2 at 00C and then transferred to a 350C water bath to induce sickling. They were embedded by the tannic acid method described below. For preparation of hemolysates, the erythrocytes were lysed with distilled water, and the membranes were spun out by centrifugation at 27,000 X g. Small volumes of 10% NaCl were added to precipitate residual membranes. After final centrifugation, 1/50th vol of streptomycin solution (5 mg/ml) was added to 1140
Biochemistry: Garrell et al. Proc. Natl. Acad. Sci. USA 76 (1979) 1141 'L 207e-I I.-so FIG. 1. Electron micrograph from a 600-A-thick section from a tannic acid-stained, embedded sickled cell. The micrograph, recorded at a goniometer stage tilt angle of 00, reveals small square packed fiber arrays in near cross section. Typically observed crosssectional patterns that arise from viewing the fibers at small angles off perpendicular to the longitudinal fiber axes include: B, bull's-eye; S, spiral; C, "C"-shaped element radiating from an edge; I, incomplete bull's-eye; U, uncoiling spiral. Scale = 0.1,um. prevent bacterial growth. The Hb S solutions were concentrated by vacuum dialysis to a final concentration of 18 mm heme and subsequently dialyzed against 0.05 M phosphate buffer (ph 6.8). Aliquots (1-ml) were stored in liquid nitrogen until required. For preparation of fiber bundles, samples were removed from liquid nitrogen and made 5 mm in IHP and 10 mm in EDTA. They were thoroughly deoxygenated by rotating slowly at 00C under N2. Polymerization was accomplished by stirring at either 20 or 370C for up to 5 min (5, 7, 8); agitation was continued until maximal turbidity was obtained (roughly 15 min). At this point, one could observe fibers and small bundles by negative staining techniques of electron microscopy. Samples were left for periods of up to 10 weeks, during which large highly ordered bundles formed and ultimately a stable crystalline phase was reached. Aliquots were removed at various times for embedding. Embedding. The tannic acid method of Mizuhira and Futaesaku (6) was used. Under water-saturated prepurified N2, cells or fibers were soaked in deoxygenated 2.5% gluteraldehyde/4% tannic acid/0.05 M P04, ph 6.5, for 1 hr. The fixed samples were washed twice in 0.05 M P04 for 15 min and then stained with 1% Os04/0.05 M P04, ph 6.5. After two 15-min changes of the phosphate buffer, samples were dehydrated with acetone and embedded in Epon/Araldite. Some hemolysate samples were centrifuged at 2000 X g for 45 min in a Sorvall S-34 centrifuge before fixation. This caused no damage to the fibers or bundles, yet served to orient them such that thin sections taken at appropriate angles contained large numbers of cross-sectioned fibers and fiber arrays. FIG. 2. Electron micrograph from a 600-A-thick section from a tannic acid-stained, embedded stirred hemolysate. The micrograph was recorded at a goniometer stage tilt angle of 420. In general, hemolysate sections reveal enhanced fiber substructure compared to sectioned sickled cells. Some of the cross sections appear in distinct patterns which have been classified; letter designations are as in Fig. 1 and: F, four-unit core; T, three-unit core. Scale = 0.1,gm. Electron Microscopy. Sections were mounted on grids, stained with methanolic 2% uranyl acetate and Reynolds concentrated lead citrate, and shadowed with a thin carbon coating to minimize electron beam damage. Specimens were examined in a Philips 301 electron microscope fitted with a goniometer stage. Computing. A Data General Eclipse system with dual disk drive and Tektronics 4014-1 graphics terminal were utilized for the computer modeling. Computed models of the cross sections presented here were photographed directly from the graphics terminal screen. RESULTS Observations on Cross Sections of Hb S Fibers. When samples were prepared for sectioning by using the tannic acid procedure, cross sections were obtained with substantial enhancement in detail compared to sections prepared by conventional methods. Electron micrographs are presented in Fig. 1 for a sectioned sickled cell and in Fig. 2 for a sectioned preparation of fiber bundles prepared from a stirred hemolysate. The hemolysate gave crisper detail and, in general, more satisfactory results have been obtained with hemolysates than with intact cells. This difference may be due to incomplete penetration of the tannic acid into the interior of the intact cells or to some interfering material in intact cells but absent in he-
1142 Biochemistry: Garrell et al. Proc. Natl. Acad. Sci. USA 76 (1979) FIG. 3. Computer models of fiber cross sections with no or small deviations of the projection plane from perpendicularity with the fiber axis. Increasing brightness levels (4) indicate increased protein density. (a) A 60-A section perpendicular to the fiber axis, showing 14 strands (Hb S molecules idealized as solid circles of equal intensity). z = 1; 0 = 00; X = 00. (b) Type F; eight 60-A sections superimposed, each rotated 7.20 with respect to the section breadth due to the fiber's helical twist (this represents the "ideal" 480-A section one would expect to observe in micrographs). z = 8; 0 = 00; 4) = 00. (c) Twelve 60-A sections superimposed as in b, to show the effect of increasing section thickness. The inner region shows a nearly closed ring because more of the helical repeat is projected than in b. z = 12; 0 = 00; 4 = 00. (d) Ten 60-A sections superimposed with the projected image tilted 10 from perpendicularity. The appearance is distinctly different from the "ideal" section in b. z = 10; 0 = 10; 4) = 0. (e) Effect of changing X, at0 5# 0. The image is rotated, and some shifts in observed intensities occur. z = 10; 0 = 10; 4 = 300. (f) Increasing 0 to 20 further alters the fiber image; the inner ring is almost complete, and a single core filament is observed, compared with four cores in the "ideal" section in b. z = 10; 0 = 20; 4 = 300. molysates. In the micrograph of a sectioned hemolysate (Fig. 2), substructure is clearly evident in the fiber cross sections and a number of well-defined patterns are observed. These patterns have been characterized in detail (see below) and, although the sections from embedded cells are less satisfactory, it is possible to identify similar patterns in them (Fig. 1). In addition, the apparent diameters of the fibers in both types of preparations have been extensively tabulated. When measurements were made of the cross-sectional center-to-center distances between adjacent fibers in close contact, either in small clusters or larger arrays, mean (+SD) values of 205 ± 5 A were obtained for tannic acid-embedded cells and 212 ± 8 A for tannic acid-embedded stirred hemolysates. Hemolysates typically contained mainly large arrays of fibers in square or hexagonal lattices, whereas cell arrays were smaller and less abundant. Nevertheless, diameter measurements in both cases were self-consistent and consistent with measurements from conventionally embedded cells and bundles (5). They also were in agreement with the diameter assessed for the 14-strand fibers observed with negative staining procedures (3), when corrections were made for flattening (5). A striking feature of the cross sections, particularly for the embedded bundles from stirred hemolysates (Fig. 2), was the presence of substructure. A number of characteristic crosssectional patterns were observed, and we have distinguished seven recurring patterns (each of which is signified by a oneletter abbreviation used to designate the pattern): (i) Four small circular regions in the core surrounded by a continuous ring; designated "F" for "four." (ii) Three small circular regions in the core surrounded by a continuous ring; designated "T" for "three." (iii) A single prominent core region surrounded by a continuous ring; designated "B" for "bull's-eye." (iv) A definite spiral appearance; designated "S" for "spiral." (v) A series of "C"-shaped elements radiating from an edge; designated "C." (vi) Resembling the bull's-eye form but lacking approximately half of the outer ring; designated "I" for "incomplete bull's-eye." (vii) Resembling the spiral form but the outer end of the spiral appears slightly separated from the internal portion of the cross section; designated "U" for "uncoiling spiral." The seven types of cross sections represent a somewhat arbitrary division of the observed patterns. In fact, there is a continuity of patterns, and certain individual patterns in Fig. 2 identified with one type also display features that would permit satisfactory assignment as another type. The main point of the assignments is to show that all of the various cross-sectional patterns can be related to projections of the 14-strand fibers as demonstrated by computer models of the cross sections. Computer Models of the Cross Sections of Hb S Fibers. To determine if the characteristic patterns observed in the cross sections of Hb S fibers embedded by the tannic acid procedure were consistent with the 14-strand model of the fibers reported by Dykes et al. (3), a computer model was developed for cross sections of 14-stranded fibers of varying thicknesses. The model was constructed to take into account small deviations of the cross sections from perpendicularity to the longitudinal fiber axis, because the fibers occur at random orientations in the embedded block and the angle at which the sections are cut will coincide exactly with the perpendicular in only a small fraction of cases. Large deviations from perpendicularity were corrected approximately with the goniometer stage of the electron microscope, but these corrections were based on visual inspection and were generally reliable to only a few degrees.
Biochemistry: Garrell et al. Proc. Natl. Acad. Sci. USA 76 (1979) 1143 FIG. 4. Computer models of fiber cross sections with moderate deviations of the projection plane from perpendicularity with the fiber axis. All projections in this figure were computed on the basis of a thickness corresponding to z = 10 (see Fig. 3). (a) Type B; 0 = 1.60; 0 = 350. (b) Type S;O = 1.7 ; X = 0. (c) Type T;O = 1.60;4 = 125.50. (d) Type I; O= 4 ; k= 45. (e) Type U; 0 = 60; 0 = 00. (f) Type C; 0 = 10 ; 0 = 450. Panels d-f show increased distortions of the fiber ellipse with increasing 0. All of these types are apparent in Figs. 1 and 2. The various patterns predicted by the computer models are presented in Figs. 3 and 4. The models in Fig. 3 correspond to fibers of varying thickness and with no or only small deviations from perpendicularity. Fig. 3a represents a 60-A section taken exactly perpendicular to the longitudinal fiber axis (z). This thickness represents the molecular spacing along each filament in the 14-strand structure. For modeling purposes, molecules are represented as solid circles of equal intensity. Because typical thin sections were between 450 and 750 A thick, 8-12 such 60-A slices or z-sections must be superimposed, each rotated 7.20 with respect to the adjacent section due to the helical twist (because 60 A is 1/50th of the full helical repeat distance of 3000 A, each section is rotated 3600/50 or 7.2 ). Model projections corresponding to 480 and 720 A are presented in Fig. 3 b and c. These thicknesses correspond to 8 and 12 molecular diameters, respectively, and illustrate the effects of section depth. Increasing the thickness of the sections results in an intensified and more continuous inner core region because more of the helical repeat is projected. The pattern presented in Fig. 3b closely resembles the F form noted in the electron micrographs of the cross sections of Hb S fibers. For subsequent patterns (Figs. 3 d-f and 4 a-f), models corresponding to an average thickness, 600 A (z = 10), were used. For models corresponding to cross sections that are not exactly perpendicular to the fiber axis, a cylindrical coordinate system was used to describe angular deviations. Deviations from the z axis are defined by the angle 0, and rotation of the 0-vector about the z axis is defined by the angle k. Small values of 0 (1-20),can lead to appreciable changes in the appearance of the model cross sections, as indicated in Fig. 3 d-f. The models of the cross sections are less sensitive to changes in 0 but, for 0 = 10, a significant change can be effected by a 300 change in A, as seen by comparing Fig. 3 d and e. As the value of 0 increases, the projection patterns become more sensitive to the value of 0. The other six forms described in relation to the cross sections observed by electron microscopy can also be generated by the computer model by selecting certain values of 0 and 0; these computed cross-sectional projections are presented in Fig. 4. With values of 0 in the range of 1-2, patterns that closely resemble B, S, and T forms are generated (Fig. 4 a-c). At higher values of 0, in the range 4-10, more severe distortions are produced and patterns that closely resemble I, U, and C forms are generated (Fig. 4 d-f). DISCUSSION Use of the tannic acid embedding procedure of Mizuhira and Futaesaku (6) results in substantial enhancement in the detail visible in cross sections of fibers of Hb S. The diameters of these fibers are consistent with the 14-strand structure for the fibers deduced by Dykes et al. (3) and are in agreement with other recent measurements of diameters from cross sections of embedded material prepared without tannic acid (5). What is of particular interest in the sections prepared with the tannic acid procedure is the appearance of substructure in the cross sections of the Hb S fibers. Seven characteristic patterns can be defined; they are especially clear in cross sections from fiber arrays produced from stirred hemolysates (Fig. 2), and several of the most prominent forms can also be detected in cross sections from embedded intact cells (Fig. 1). The various patterns correspond to projections of the 14-strand structure at various angular deviations of the section plane from perpendicularity with the fiber axis, as shown in Figs. 3 and 4. Thus, the data from embedded and sectioned fibers support the prevalence of Hb S fibers in the 14-strand form not only in terms of the agreement in diameters but also in terms of the correspondence between substructure patterns in the cross sections and predictions for such sections derived from the 14-strand structure. A remaining issue is the uniqueness of the 14-strand model in representing the cross-sectional images. No alternative models were found that satisfactorily represented cross sections, but efforts in this regard were not exhaustive. Thus, we are not asserting a proof for the 14-strand structure from the data presented here but simply wish, at this stage, to demonstrate a compatibility. A more extensive study will be required to rule out alternative models conclusively and such a study will also have to consider issues such as possible nonuniform stain pen-
1144 Biochemistry: Garrell et al. FIG. 5. Model of fiber array, illustrating the effect of tilting 300 with the goniometer stage. As the section is tilted to restore the bundle cross section to near perpendicularity with the fiber axes, projected images of adjacent fibers differ in angular orientation because sectioning has cut staggered portions of the fiber helices. The consequence of this effect, when restoration of perpendicularity is imperfect (0 0), is the projection of images of adjacent fibers corresponding to differing values of 0. etration through the sections and the size of the individual disks used to represent Hb S molecules (as in Fig. 3a). Concerning this latter point, effective disk size presumably is smaller than the actual molecular diameter, as a result of positive staining effects. In the present study, disk sizes were varied arbitrarily to maximize agreement with the cross-sectional images. The ability of the tannic acid procedure to enhance the detail of the Hb S fiber cross sections to such a significant degree indicates that its effects are not specifically related to the properties of microtubules but have a more general basis. Such a generality would support the mode of action of tannic acid proposed by Fujiwara and Tilney (7). In their view, tannic acid molecules act by attaching to and surrounding protein units by hydrophobic and hydrophilic interactions. The multiple hydroxyl groups of tannic acid then react with the OS04 prestain and the uranyl acetate and lead citrate poststains, providing an electron-dense matrix excluding the actual protein units. The result is essentially a negative staining of the protein units, compared to positive staining effects that prevail in the absence of tannic acid. Resolution is increased in the same manner that Proc. Natl. Acad. Sci. USA 76 (1979) negative staining procedures generally produce higher resolution than do positive staining procedures. The next higher level of analysis of the cross-sectional patterns of the Hb S fibers would be to correlate the patterns of neighboring fibers in arrays in order to obtain information on interactions between fibers. If neighboring fibers were in register, in terms of the long and short axes of their roughly elliptical cross sections (3), identical cross-sectional patterns might be expected. The fact that most micrographs in our study were recorded at finite goniometer angles, however, in an effort to correct for deviations of the section plane from perpendicularity to the fiber axis, introduces an effective staggering that is equivalent to substantial differences in k between adjacent fibers. This aspect of the analysis is summarized in Fig. 5. For example, for fibers in a square packed array tilted 300 to correct for deviations of the section plane from perpendicularity with the fiber axis, adjacent fibers are staggered along the z axis by about 100 A. This would correpond to an effective difference in k of 120 (100/3000 X 3600) for adjacent fibers and would result in significantly different cross-sectional patterns for adjacent fibers for 0 # 0 (as is almost invariably the case). It may be possible to reconcile an array of fibers with values of 0,k and a given tilt angle for the goniometer stage, but a more complicated modeling program would be required. We thank Dr. John Telford for help with the sectioning and Dr. G. Dykes, E. Fram, and C. Akey for helpful discussions. This work was supported by grants from the National Science Foundation (PCM 76-16760), the National Institutes of Health (CA-14454), and the National Foundation-March of Dimes (1-611). 1. Finch, J. T., Perutz, M. F., Bertles, J. F. & Dobler, J. (1973) Proc. Natl. Acad. Sci. USA 70,718-722. 2. Ohtsuki, M., White, S. L., Zeitler, E., Wellems, T. E., Fuller, S. D., Zwick, M., Makinen, M. & Sigler, P. B. (1977) Proc. Natl. Acad. Sci. USA 74,5538-5542. 3. Dykes, G., Crepeau, R. H. & Edelstein, S. J. (1978) Nature (London) 272,506-510. 4. Dykes, G., Crepeau, R. H. & Edelstein, S. J. (1979) J. Mol. Biol., in press. 5. Crepeau, R. H., Dykes, G., Garrell, R. L. & Edelstein, S. J. (1978) Nature (London) 274, 616-617. 6. Mizuhira, V. & Futaesaku, Y. (1971) Annu. Proc. Electron Microsc. Soc. Am. 29,494-495. 7. Fujiwara, K. & Tilney, L. (1975) Ann. N. Y. Acad. Sci. 253,27-50. 8. Pumphrey, J. G. & Steinhardt, J. (1976) Biochem. Biophys. Res. Commun. 69,99-105.