Preparation of Biological Tissues for Electron Microscopy by Freeze-dryi ng 5 H. DAVID COULTER AND LOUIS TERRACI03 Department of A n a t o m y , Lrnruerstty of blrnnesotcz. Mznneapolzs, Minnesota 55455 ABSTRACT A dependable method for freeze-drying tissues for electron microscopy has been developed. Thin slices of fresh tissue were frozen by bringing them into direct contact with a polished copper bar at liquid nitrogen temperature. The tissue was transferred to a copper specimen block equipped with a thermocouple and heating circuit for accurate control of the environmental temperature of the tissue, and evacuated i n a glass freeze-drier using clean high vacuum techniques for keeping the system free of hydrocarbons. The tissue was dried by increasing the temperature of the specimen block 1 0 ° C each hour while monitoring the rate of water removal from the tissue with a partial pressure analyzer. The dry tissue was fixed with Os04 vapor, vacuum embedded i n a low viscosity epoxy resin, sectioned, stained, and viewed with the electron microscope. Tissue processed i n this manner exhibits excellent morphological preservation at both cellular and organellar levels without prefixation or the use of cryoprotective agents. The results of the experiments using the partial pressure analyzer indicate that small blocks of tissue can be dried i n a short time at low temperature. Intermittent but serious attempts to freeze-dry biological tissue for electron microscopy have been made since the earliest years of electron microscopy (Richards et al., '42, '43; Sjostrand, '43). At that time most electron microscopists were concerned primarily with obtaining presentable micrographs, and freeze-drying was one of many possible methods worthy of exploration. Although the micrographs obtained were as good as many others that appeared in the literature at that time, it gradually became apparent that fixation in aqueous solutions was superior for general application, and interest in freeze-drying faded as electron microscopic cytology developed in the 1950s. More recent efforts to freeze-dry tissue for electron microscopy have utilized the technique for specialized applications, such as radioautography of soluble tracers (Stirling and Kinter, '67), electron microscopic cytochemistry (Gersh, '73), and maintenance of cellular relationships thought to be distorted by aqueous fixation (Van Harreveld and Malhotra, '66).The quality of preservation exhibited has always been poor, however, when compared with standard methods, and methodological difficulties have been serious and often intractable. For ANAT. REC., 187: 477-494 these reasons, the technique has never come into routine use despite its theoretical appeal. Satisfactory freeze-drying of tissues for electron microscopy requires first, that the tissue be frozen in such a way that ice crystal formation is minimized, and second, that the tissue be dried at low temperature. This paper presents in detail a technique for obtaining well frozen material and for drying it in a controlled environment. The superiority and efficacy of the method is proven by the most important morphologic standard, the excellent preservation of fine structure. Additionally, data on drying rates and theoretical considerations of the drying process make it clear that drying can be accomplished in a short period of time, and that a very simple and inexpensive drier (Terracio and Coulter, '75) is all that is needed for good results. Received June 15, ' 7 6 . Accepted Oct. 14, '76. 'Work o n this technique was initiated at t h e University of Tennessee Medical Units, where i t was supported by USPHS Training Grant GM 00202. At the University of Minnesota, this work was supported by the Minnesota Medical Foundation. the Graduate School, and USPHS Training Grant GM114. Dr. Terracio's present address is the Department of Anatomy, Hahnemann Medical College, Philadelphia. Pennsylvania 19102. 477 4 78 H. DAVID COULTER AND LOUIS TERRACIO MATERIALS AND METHODS Freezing method This freezing technique was developed from a method described first by Van Harreveld and Crowell ('64). A copper bar 8 inches high by 3 inches in diameter with a mirror finish on one end was polished with jeweler's rouge, rinsed with xylene, and dried gently with a flannel cloth, The bar was placed in a 10-liter Styrofoam ice chest fitted with a lid containing a centrally placed 3-inch diameter hole. The mirror surface of the bar was about 2 inches below the lid of the ice chest (fig. 1). The ice chest was filled with liquid nitrogen and kept full for two and one-half minutes. Approximately 11 more minutes were required for the bar to reach liquid nitrogen temperature, at which time the level of the liquid nitrogen was 2 inches below the freezing surface. Adding nitrogen to the container after that time was avoided because it remains on the upper surface of the bar if it is added after the bar is chilled and possibly retards freezing by limiting direct contact of the tissue with the copper surface. Fresh tissue was cut by hand with a razor blade into slices about 1/4 mm in thickness and were placed on a No. 5 cork stopper covered on its small end with thin gauge (0.0007 in) aluminum foil. The stopper was grasped with a pair of Russian tissue forceps so that the surface of the tissue remained perpendicular to the long axis of the forceps. The tissue was blotted gently to remove excess blood and promptly frozen by bringing it into direct contact with the surface of the bar and holdi n g it firmly for 20 seconds. If the tissue did not meet the bar squarely, or if it slipped or rotated on the bar after the initial contact, the tissue was discarded. After 20 seconds the aluminum foil with its adhering tissue was slipped from the stopper and dropped into the surrounding liquid nitrogen. It was possible to freeze 20 or more slices of tissue in this manner on the 3inch diameter surface before having to warm and repolish the bar. The area where each piece of tissue had been frozen was clearly visible, arid care was always taken to freeze only on a clean area of the surface. When a sufficient amount of tissue had been frozen, it was removed from the aluminum foil and transferred to a 10 ml Fig. 1 Styrofoam ice chest (about 7 X 9 X 10 inj c u t away to show copper bar. beaker immersed in liquid nitrogen. The tissue was then dissected into pieces about 1 mm square by chipping i t with dissecting needles chilled with liquid notrogen. Drying method The freeze-drying apparatus (fig. 2 ) was constructed and designed to minimize the accumulation of hydrocarbons and to achieve a vacuum in the torr range. Both of these features are necessary for analysis of drying rates. Figure 2 shows the freeze-drier in the prepurnping stage of operation. It was supported at such a height that 7-inch O.D. glass Dewars could conveniently be lifted into place under the specimen tube (E) and pumping chamber (K). Supporting hardware, clamps for the O-ring joints, electrical connections, mechanical pump, and large glass Dewars for the pumping tube and specimen tube are not shown. Starting with the system at atmospheric and room temperature, the pumping chamber (K) was preheated at least 12 hours at 150" C. With N r blowing through the system, the specimen tube (E) was lowered allowing the specimen block (G) to hang below the rest of the system supported only by the heater wires. The specimen block was brought to liquid nitrogen temperature by surrounding it with a small, stable container of liquid nitrogen. FREEZE-DRYING OF TISSUES FOR ELECTRON MICROSCOPY The level of liquid nitrogen around the specimen block was maintained carefully after equilibration so that it just covered the surface, and the small pieces of frozen, chipped tissue were transferred to it with chilled fine forceps. The actual transfer of tissue from the beaker to the circular depressions in the specimen block was effected over as short a distance and as quickly as possible. Nitrogen in liquid form was visible on the blocks during transfer. The tissue was always kept immersed in liquid nitrogen at other times to obviate concern about heat transfer directly from atmosphere to the specimens. The bottom one-fourth of the specimen tube (E) was precooled for about 30 seconds with liquid nitrogen, the container of liquid nitrogen for the specimen block was lowered, and the specimen tube was promptly attached with an O-ring clamp to the freezedrier, A small Dewar (H) with liquid nitrogen was then placed around the bottom of the specimen tube to maintain the copper block temperature for the next one to two hours. Next, in a sequence of three steps, the mechanical pump (not shown in the diagram) was connected to the system by way of the U-tube (B), the pumping chamber (K) was removed from the oven and connected to the system with an O-ring and clamp, and the bottom 3 4 inches of the U-tube was immersed in liquid nitrogen (Dewar C ) for a few seconds. Then the mechanical pump was started. As the system pumped down over the next 30 minutes, liquid nitrogen was gradually added to the Dewar (C) until it was nearly full. A considerable volume of water came off the hot molecular sieve (K), and if the Dewar trap (C) was filled immediately, water sometimes blocked the U-tube completely on the upper right hand side. If the Dewar was filled gradually, water was trapped over the entire extent of the right side of the tube, and it remained patent. Maintaining the system in this way, with nitrogen in both Dewars (C and H), the vacuum dropped to about 5 X lo-' torr in an hour, the temperature of the block rose to about - 160"C, and the molecular sieve pump gradually cooled off. Valve A was then closed and the system appeared as shown in figure 2. The mechanical pump and the U-tube trap were now removed from the system and a large glass Dewar (7-inch O.D.) was 479 placed around the pumping chamber (K) and filled with liquid nitrogen. A vacuum of about 1 X torr was achieved in about two hours. The small Dewar (H) around the specimen tube was then replaced with a large Dewar, the partial pressure analyzer was turned on, and the entire system was allowed to equilibrate overnight . The copper block achieved a steady state temperature of about - 130"C, the total pressure dropped to approximately 1 X 10 torr, and the partial pressure of water torr. The temperareached about 6 X ture of the copper block was then raised 10°C per hour until it reached room temperature, and the concomitant changes in1 the partial pressure of water were recorded on a strip chart. For achievement of reproducible results on drying rates it was necessary to maintain the levels of liquid nitrogen in both Dewars for the duration of the experiment. This precaution was not necessary for routine use of the system for freeze-drying without the use of the partial pressure analyzer. Fixation a n d embedding After the copper block reached 2 5 ° C the partial pressure analyzer was turned off and the pressure in the system was brought to atmosphere through the isolation valve (fig. 2 : A) with dry nitrogen. The specimen tube was lowered and the dry tissue was immediately transferred to a small glass fixing chamber (fig. 3). This chamber had been prepared earlier. With valves B and D closed, a molecular sieve pump had been attached to the evacuation valve (D) and chilled for two hours with liquid nitrogen. This pumping chamber was used only for this purpose and was somewhat smaller than the one for the freeze-drier. After transferring the tissue to the fixing chamber the lid was securely clamped into place and valve D was opened cautiously. To evacuate the atmospheric gases from the tube below valve B and yet retain the fixative, tube (A) was first chilled with liquid nitrogen for a few minutes. Then valve B was opened, the evacuation valve (D) was closed, and tube (A) was allowed to come to room temperature. The tissue was fixed 4 to 12 hours with OsOl vapor. Following fixation valve B was closed, and the pressure in the chamber was brought to atmo- 480 H. DAVID COULTER A N D LOUIS TERRACIO itJ Fig. 2 Freeze-drying apparatus i n prepumping stage, one-fifth actual size. A high vacuum bakeable glass valve (Kontes, Vineland, New Jersey 80360) with 17-mm O.D. side arms, teflon plunger, a n d Viton O-rings (A) isolates the freeze-drier from a liquid nitrogen trapped mechanical pump, a n d also serves a s a n up-to-atmosphere valve. A 17-mm O.D. U-tube (B) with a n O-ring joint for coupling to the valve serves a s a trap for the mechanical pump. A Dewar flask (C) chills the U-tube with liquid nitrogen. Heater wires a n d thermocouple wires from a temperature controller that can regulate temperatures to - 196OC (Leeds and Northrup, North Wales, Pennsylvania 19454) enter the system at D. Short segments of ceramic thermocouple tubes (Omega Engineering, Stamford, Connecticut 06907) were glued into holes bored in a Kovar closed-end cap (Latronics Carp., Latrobe. Pennsylvania 15650) with epoxy glue. The heater a n d thermocouple wires were then threaded through the ceramic tubes a n d adjusted to a height that would barely allow the specimen block (G) to rest on the bottom of the specimen tube ( E j . They were then glued into place at that height. The specimen tube (E) is made from 4 5 m m O.D. Pyrex tubing, closed off and flattened at t h e bottom a n d fused to a n O-ring joint at the top. The thermocouple wires and heater wires a r e shown at F. The thermocouple wires are 0.010-inch bare copper and constantan wires (Omega Engineering) .joined with a bead that is snugly fitted into the specimen block. (If it is not tightly fitted, the temperature regulation will be erratic). The thermocouple bead a n d the hole for accommodating it are disproportionately large in the drawing. The heater wires that pass through the ceramic tubes are 0.018-inch copper. They are attached to fiber glass insulated wires t h a t connect to the cartridge heater i n the specimen block. The copper specimen block (GI is 1 by 2 314 inches with a central li4-inch hole. sized precisely for accommodating a tightly fitted 100-watt cartridge heater 1/4 inch from the bottom of the hole. One-fourth inch holes are bored all around the periphery of the specimen block for accommodating the specimens. The pumping chamber (K) is made from 100 m m Pyrex tubing a n d contains 5 A molecular sieve. The thermocouple gauge (I) a n d partial pressure analyzer (J) (Ultek Division, Perkin-Elmer, Palo Alto, California 94303) are mounted with standard ultra-high vacuum flanges to the rest of the system by means of a glass to Kovar seal. FREEZE-DRYING OF TISSUES FOR ELECTRON MICROSCOPY n 48 1 infiltration floated back to the surface, and the others were properly infiltrated at the end of six hours. They were usually oriented at a n angle in the bottom of the capsules, and the surface which had come into direct contact with the bar during freezing was easy to locate and preserve during sectioning. RESULTS b'reezing The gross appearance of the tissue after freezing depended upon the force with which the tissue struck the bar, the type of tissue, and the manner in which it was sliced. A 114 mm slice of highly parenchymatous tissue such as kidney or liver brought down firmly against the copper bar had about the same size and shape as before freezing. Tissues which were difficult to slice evenly and which did not retain their shape after slicing such as brain or intestine were flattened. Also, a Fig. 3 Fixing chamber. Os04 is shown at A small cube of tissue was more likely to i n the bottom of a 7-mm tube fused to a n O-ring joint. A small, high vacuum glass valve (Kontes) become pressed into a sheet. A n impression isolates the Os04 from the fixing chamber (C) of the tissue was left on the bar in the except during fixation. The fixing chamber w a s form of a clear area surrounded by a made from a 45-mm O-ring joint, flattened and small area of frost condensation, providclosed on the bottom, and is prepumped through anotherhigh vacuum glassvalve (D) by a molecular ing an immediate record of whether it sieve pump which is used only for this purpose. has been properly frozen. sphere. OsO, was left in the tube at A and used repeatedly. After vapor fixation the tissue was vacuum embedded in a low viscosity epoxy resin (Spurr, '69). The resin was mixed carefully to avoid excess bubbles, placed in polyethylene capsules, and degassed in a glass desiccator with a water aspirator or a positive displacement pump. For infiltrating the tissue, a single block of tissue was placed on the surface of the degassed resin in each capsule, and the desiccator was evacuated cautiously to avoid bubbling. As soon as the pump reached equilibrium, it was removed and the tissue was allowed to infiltrate under vacuum for two to six hours. Completion of infiltration was indicated by the sinking of the blocks in the resin. If some of the blocks did not sink in two hours, the system was brought to atmosphere, the blocks were mechanically pushed into the resin, and the desiccator was re-evacuated. Blocks which were still too large for good Drying Drying characteristics of a group of about 100, 1/4 X 1 X 1 mm blocks of kidney are shown in figure 4. For the first four hours, each time the copper block temperature was raised by 10°C, the partial pressure of water in the system rose sharply and then declined gradually. Also, the maximum peak heights declined each hour. These sharp peaks and characteristic declines ended abruptly at - 8 0 ° C . From - 70°C to room temperature, the partial pressure of water remained near the base level, indicating that no further water was being sublimed from the tissue after the first five hours of drying. The blocks of tissue retained their shape during the drying process. They had a vaguely fluffy appearance when fully dry at room temperature but were not visibly distorted or shrunken. Much larger blocks became misshapen and distorted, even when long drying times were employed. 482 11. DAVID COULTER AND LOUIS TERRACIO I lo 9 8 7 I- LL 3 3 2 I TIME (HRS.) I 0 DEGREES(C1 -120 I -110 2 3 -100 -90 4 5 6 7 8 9 10 -80 -70 -60 -50 -40 -30 -20 II 12 -10 O Fig. 4 Partial pressure analysis of the drying time for 114 mni blocks of kidney. At time 0. after pumping the system overnight with the molecular sieve pump. water w a s equilibrated at about G X 1 0 - 9 torr. a n d the copper block temperature was about -130OC. Each time the temperature of the copper block was raised by lO"C, the partial pressure of water i n the system rose sharply, well into the 10-8 torr range, a n d then declined to about 1 X l o - * torr. A n imaginary line drawn from point to point at the top of each peak would also show a decline, with a rather sudden termination of the s h a r p peaks at -80°C. The tissue was considered essentially dry after about five hours. when the copper block temperature w a s raised to - 70°C . Fixation arid e?nbcdding During vapor fixation with OsO, the tissue developed a slightly greyish cast. This greyish appearance was slightly more pronounced on the side of the tissue that came into initial contact with the copper bar. A smoothness, almost a glossiness, was apparent on this surface which was in clear contrast to the rougher appearance on the side of the tissue which had been in contact with the aluminum foil during the freezing procedure. Large, shrunken blocks turned black during vapor fixation rather than grey. Upon infiltration, the grey tissue turned completely black. The tiniest slivers of tissue became infiltrated and turned black the moment they came into contact with the resin. The other blocks turned black as they sank into the resin during vacuum infiltration. Some osmium-containing materials leached from the blocks during infiltration, especially when the longer periods of vapor fixation had been employed. Light microscopy Inspection of the frozen-dried tissue with the light microscope revealed whether or not the material would be useful for electron microscopy. If ice crystals were visible in 1/2 Fm sections under oil immersion near the surface of the block that first contacted the bar, the material was invariably unsatisfactory for electron microscopy. In blocks of tissue that had been FREEZE-DRYING OF TISSUES FOR ELECTRON MICROSCOPY 483 block, and no drying or compression artifacts of the type discussed above are present, Electron microscopy Fig. 5 One-fourth micron section of fiozendried mouse kidney. No ice crystal formation. compression artifact. or gross artifacts of drying a r e present. X 1,000. well frozen, that contacted the bar squarely and remained in position without subsequent movement, ice crystals were found on the far surface of the tissue and disappeared as one scanned toward the side that had been frozen initially. If the block of tissue had slipped or rotated on the bar during freezing, the cellular integrity was disrupted, and ice and compression artifacts were evident throughout the block. Attempting to dry blocks which were too thick yielded characteristic artifacts. The blocks were not completely infiltrated if they were thicker than 2 mm, and blocks in the 1-2 m m range often contained a core of densely stained, macerated and unrecognizable material which was surrounded on all sides by intact tissue which had been satisfactorily dried. Figure 5 illustrates a 1/4 pm section of frozen-dried mouse kidney. A small arteriole is shown in the center of the field, surrounded by kidney tubules. Nuclei, mitochondria, and cytoplasmic vacuoles are all prominent, Occasional breaks in the tissue between tubules probably occurred during drying or subsequent handling. No ice crystal formation is evident, even at the bottom of the print, 180 p m into the Tissue processed by this method yields excellent morphological preservation (figs. 6-9). Except for obvious shrinkage spaces, morphological relationships are intact at all levels of organization, tissue, cellular, and subcellular. Cells retain their normal shapes, and plasma membranes and intracellular membranes are always intact. The inherent contrast of the material is remarkably high. Membranes have nearly as much contrast in unstained frozen-dried sections that have been vapor-fixed with OsO4 as they do in double-stained aqueousfixed specimens. Two types of characteristic artifacts were always present, ice crystal damage and shrinkage spaces. The ice crystal reticulations were very predictable, increasing in size with increasing distance from the surface which was initially frozen. Easily recognizable shrinkage spaces occurred most commonly around mitochondria. Figure 6 is a micrograph of renal proximal tubule exhibiting excellent morphological preservation. No discontinuities are seen in plasma membranes or intracellular membranes. The basal infoldings, microvilli, and Golgi apparatus all appear normal. The mitochondria are of special interest. Two clearly defined populations are found. One group of very dark mitochondria has extremely electron dense matrices in which the cristae are barely discernible, and always seem to be surrounded by prominent shrinkage spaces. A smaller group of mitochondria has light matrices with concomitantly prominent cristae, and are seldom if ever surrounded by shrinkage spaces. At this magnification, cytoplasmic reticulations due to ice crystal growth become evident in the upper portion of the cell, which is several microns from the surface initially frozen by the bar. Inexplicably, they are also found in the basement membrane, which is only about 112 pm from the surface of the block. Figure 7 is a micrograph of surface epithelial cells of rabbit gallbladder, frozen by bringing the luminal surface into contact with the copper bar. A characteristic lateral intercellular space is seen between 484 H. DAVID COULTER AND LOUIS TERRACIO Fig. 6 Renal proximal tubule cells. frozen-dried, fixed i n O s 0 4 vapor, stained bv uranvl acetate arid lead citrate. Basal infoldings are seen below and microvilli at the top of the print. Most of the mitochondria have a dense matrix (DM), and are surrounded by a shrinkage space. A few others have a light matrix (LM), and lack a shrinkage space. Many of them are connected with one another by small buds (light arrow). An array of 30-50 nni electron dense reticulations due to ice crystal growth are seen in the cytoplasm at the top of the print a n d in the basement membrane below (dark arrows). x 22.000. FREEZE-DRYING OF TISSUES FOR ELECTRON MICROSCOPY 485 Fig. 7 Rabbit gallbladder, luminal epithelial c emils, frozen-dried, fixed by OsO, vapor, stained by uranyl acetate and lead citrate. X 27,000. adjoining cells, The mitochondria shown in this particular micrograph are almost all of the dark matrix variety. Figure 8 illustrates the circumstance that commonly occurs at the edge of the block. The nucleus of an unidentified kidney cell is at the interface that was initially frozen. Nuclei seem to be especially susceptible to ice crystal formation. No de- tectable ice crystal growth occurs at the initially frozen surface, down to the limit of resolution in this micrograph of about 2 nm. Extremely small ice crystal reticulations occur on the other side of the nucleus. The section was unstained, illustrating the high inherent contrast of vapor fixed blocks. Figure 9 shows a high resolution stereo- 486 €1. DAVID COULTER AND LOUIS TERRACIO Fig. 8 Frozen-dried kidney. fixed by Os04 vapor, n o additional staining. The cytoplasm of the cell in the lower part of the print is disrupted because the razor blade passed directly through the cell. .just missing the niicleus and allowing the contents of the cell to spill away before it was brought into contact with the bar. X 30.500 scopic view of renal proximal tubule cells contacting basement membrane. The inner lamina of the trilaminar membrane is especially prominent, The basement membrane shown on the left is only about 200 nm from the surface that was initially frozen. The section was about 30 n m in thickness, and the field that is shown here in stereo is about 300 nm in width, The 10:1 ratio of width to depth accounts for the limited appreciation of depth perception viewing the micrographs with a stereo viewer. No evidence for ice crystal reticulation is evident, even in the basement membrane, where ice crystal formation often seems especially prominent, DISCUSSION Successful execution of this technique requires meticulous attention to detail; even slight modification of the technique may affect the results adversely. Frwzing technique The freezing operation is the step that always limits quality if the tissue is dried, fixed, and embedded according to the protocols outlined in this paper. The crucial moment is the initial contact of the tissue with the bar and the maintenance of that contact for 20 seconds without movement. The glass transition temperature of water is about - 137°C (Rasmussen and MacKenzie, '71), and if the tissue is not brought to that temperature quickly, ice crystals will form. Experience will soon direct the operator to a reasonable balance of force and velocity in making contact with the bar, not squashing the tissue, yet not being too delicate with it. The use of a cork stopper provides a slight cushioning effect that is very helpful, absorbing minor unsteadiness and keeping the tissue placed where it initially made contact with the bar. The use of a more rigid supporting material such as a wooden dowel makes the operation much more critical. The slightest movement of the hand is transmitted to the contact area between the metal and the tissue, re- FREEZE-DRYING OF TISSUES FOR ELECTRON MICROSCOPY 487 Fig. 9 Frozeri-dried kidney. fixed by OsO, vapor, double stained by a n aqueous solution of uranyl magnesium acetate at 5OoC a n d lead citrate. The micrographs were taken at plus a n d minus 10" of tilt on a n AEI 801 at 60 kV, using a n initial magnification of X 160,000. Final magnification is X 220,000. sulting in slipping or rotation of the sample on the bar and invariably spoiling the tissue for electron microscopy. At the moment of freezing, heat is being rapidly conducted from the sample to the bar. If the process is interrupted or delayed before both the tissue arid the aluminum foil reach a low temperature, the tissue will warm up due to conduction of heat from the aluminum foil, resulting in devitrification of the water and extensive ice crystal growth. Drying technique Theoretical consideration of the variables that are important in freeze-drying of tissue blocks is a n invaluable aid to the design of a simple and inexpensive drying apparatus, and to understanding why tissue can be dried in a relatively short time. There are five major variables that must be considered, (1) the thickness and (2j the temperature of the dry shell of tissue that develops on the outer surface of the block during the drying process, ( 3 ) the initial surface area of the sample per unit of volume, (4) the temperature of water at the drying boundary between the frozen core of tissue and the surrounding shell of dry tissue, and (5) the efficiency of the drier in condensing water that escapes from the sample. Stephenson ('53) first considered in detail the occurrence and growth of a shell of dry tissue on the outer surface of frozen blocks. This shell contains all cytoplasmic materials with the exception of water and dissolved gases, and is of great importance in impeding the flux of water molecules from the interior core of frozen tissue to the surface of the block. Each molecule will make a vast number of collisions with organelles and other structures in the cytoplasm, and will be adsorbed and released at a temperature-dependent rate at 488 H . DAVID COULTER AND LOUIS TERRACIO each site, so the number of molecules that finally escape from a unit area of dry surface per second will depend on both the thickness of the shell and its temperature gradient. The most easily manipulated variable is surface area. A sheet of tissue, having a larger initial surface area than a cube of equal volume, will obviously be more easily dried, Not only will the availability of a larger initial surface area yield an increased initial drying rate, but the block will dry to completion before the shell becomes as thick as it will be for the same volume of tissue in a cube. Next, the rate of drying will increase as the temperature and vapor pressure of water in the frozen core increases, provided the increased water that is sublimed can move through the shell. As Meryman ('60) has pointed out, the equilibria that occur at the drying boundary are crucially affected by the resistance of the dry shell to vapor flow, Last, it is important that the water that does finally escape from the outer surface of the dry shell be trapped efficiently by the freeze-drier. The partial pressure analysis of drying rates has aided considerably in removing some of the unknowns from the technology of freeze-drying of tissue samples for electron microscopy, and provides a clear rationale both for tissue drying time and for temperature regulation of the copper specimen block. The base level of 6 X 10-9 torr for the partial pressure of water (fig. 4) represents a steady state value, Water is being continually desorbed from all inside surfaces of the vacuum system, and adsorbed onto them. When water is adsorbed onto a surface at liquid nitrogen temperature, such as the inside surfaces of the pumping tube and the specimen tube, or the molecular sieve, it is trapped there, or pumped, for as long as the temperature is maintained, Each time the temperature of the copper block is raised 10°C, a small fraction of the escaping water leaves the samples at such an angle that permits access to the orifice of the specimen tube, where it escapes into the rest of the system and upsets the equilibrium, The upward limb of each peak in figure 4 occurs when an increase in conductive and radiative heat from the specimen holder causes increased vaporization of water from the drying boundary in the samples and an increased flux of water to their surfaces. It is important to understand that the temperature of the water in the core of the sample may not have an obvious relationship to the temperature of the specimen holder, Increased heat is no doubt radiated from the copper block each time its temperature is raised by a 1 0 ° C increment, but the increase in temperature of water inside the specimen will depend on many factors, the most important of which is the degree to which the dry shell insulates the core. This insulating effect may be considerable, and presumably becomes more important as the thickness of the shell increases. Indeed, the argument will be developed later that the primary effect of the increasing copper block temperature is on the shell rather than on the frozen core, The descending limb of each peak (fig, 4) occurs for two reasons: first, the sublimation of water produces evaporative cooling, causing the temperature of the core of the sample to drop and thereby checking the rate of sublimation; and second, the dried shell of the sample becomes slightly thicker and more effectively insulates the frozen core from its heat source, decreasing the amount of heat available for driving sublimation at the drying boundary. To consider the drying process more rigorously, it is helpful to model the ideal drying characteristics of a single block of pure ice, 1 X 1 X 114 mm, a special case of 1 X 1 X h, finding the decrease in volume with respect to time, -dV/dt, keeping all other variables constant. No shell is present, the initial surface area (S) of 3 mm2 (212 4hl) and volume (V) of 114 mm:1 (h12) are given, the temperature of the block is kept constant with whatever heat input is necessary to counteract evaporative cooling, and the condensing process is 1 0 0 8 efficient. Water sublimes from all surfaces equally so that all dimensions of the block decrease by equal amounts as drying occurs. To find the necessary expressions, consider first that the new volume of ice after a short time of sublimation or V(t A t), will be equal to the old volume minus the water that comes off the surface area during that period of time, or V(t) - k A t S(t). As A t approaches 0, it + + 389 FREEZE-DRYING OF TISSUES FOR ELECTRON MICROSCOPY can be shown that dVldt = -kS. This simply states that volume is decreasing at a rate proportional to surface area, a relationship that is readily appreciated intuitively. At t = 0, V = hl? and S = 212 + 4hl. Since all dimensions decrease equally, h can be defined in terms of 1 throughout the course of drying as 1 -314. For example, initially 1 - 3/4 = 1/4, and at a later time when the block has lost water to a depth of 1/16 mm from all surfaces h is still 1 - 314, or in that case 718 - 3/4 = 118. Now V and S can be defined in terms of 1, not only at t = 0 but for the entire course of drying, as V = 13 - 31'/4, and S = 612 - 31. Now, stating dVldt = - kS in terms of 1, it can be shown that dl/dt = -2k, which is simply a statement that the decrease in any given dimension with respect to time is a constant, Again, this can be appreciated intuitively by understanding that the molecules of water would come off a unit of surface area at a constant rate dependent only on vapor pressure, and that the constant depletion of water from any two opposite surfaces dictates a constant decrease in the distance remaining between them. Integrating the expression dl = - 2k dt yields 1 = - 2kt C, and since at time 0, 1 = 1, it follows that C = 1, and 1 = 1 2kt. Finally, since V = 1s - 312/4, the equation showing the relationship V = f(t) is found to be a cubic with the form V = (1 - 2kt)3 - 3(1 - 2kt)'/4. The meaningful part of the curve between V = 0.25 and V = 0 is shown in figure 10. The height of the curve at any time t represents the volume of ice present at that time, and the slight change in the slope of the curve as the volume approaches 0 represents a slightly decreased rate of drying as surface area decreases. This is the pertinent function. A little introspection will make it clear that if a n ideal partial pressure analyzer were being employed to study the model, it would be measuring a function of the rate of drying, not the volume of ice remaining in the block. It actually would be measuring small decreases in volume (the molecules that are sublimed away) over small periods of time at any given moment, or the negative of AVlAt versus time. So the equation Fig. 10 V = f(t). I .5 --dvd t + t Fig. 11 0.125 -dV/dt = X t ) . needed to compare the model with the data is not V = f(t), but the negative of its first derivative, -dV/dt = f(t), or -dV/dt = 24k3t2 - 18k2t 3 k (fig. 11). Again, the curve has no physical meaning after the block is dried at t = 0.125, and the drying rate becomes zero at that time. The analysis given holds for t between to and t I where t I is the time at which the volume becomes 0 . At t , , 1 = 314, 1 2kt1 = 3/4, and k = 1/8t,. For figures 10 and 11, units of time are chosen so that t , = 1/8,making k = 1 . Examining the model further, it is + 490 H. DAVID COULTER AND LOUIS TERRACIO clear that the reason the drying rate is still quite substantial just before t = 0.125 is that there is still considerable area on the top and bottom surfaces of the block as the thickness approaches 0 . The behavior of a model of a cube of ice is quite different, with the drying rate (and surface area) decreasing gradually as all dimensions decrease identically toward 0. In that case there is no sharp cutoff of - dV/dt as drying is completed. The plot of - dVldt = f(t) shows a definite qualitative resemblance to the data shown in figure 4. Since the upward limbs for the individual peaks in figure 4 represent a n approach to steady state for each new temperature, and since the downward limbs result from evaporative cooling, a tentative comparison can be made by drawing a curve through the tops of the peaks in figure 4 and superimposing that over the curve of the model. If that is done the shapes of the curves coincide sufficiently to require explanation. Only the first two peaks in figure 4 are slightly higher than predicted from the model. The resemblance of - dV/dt = f(t) for a block of ice to the drying characteristics of the tissue blocks suggests that the five variables discussed earlier either behave as constants using the drying technique described in this paper or that they cancel each other, The third and fifth variables, in fact, are certainly constant. The efficiency of the drier in condensing water that escapes from the sample is purely a function of freeze-drier design, and the initial surface area of the blocks per unit of volume is set as soon as the blocks are prepared. Both of these variables are important, but they cannot change after the drier is built and the tissue is prepared. The first and second variables tend to cancel each other. The drying rate is decreased with increasing thickness of the shell and increased by the warming of the shell by the copper block. A temperature gradient is established in the shell, as cold as the ice at the drying boundary and as warm as the copper block on the outer surface. If, as will be discussed next, the temperature of water remains relatively constant in the core, the temperature gradient in the shell must become very steep as drying proceeds. Each water molecule is condensed on innumerable surfaces in its course across the shell, and its release in each case is dependent upon vapor pressure. The occurrence and steepness of the temperature gradient in the shell therefore becomes a n extremely important factor in facilitating vapor flow. For example, if the temperature of the ice were - 100°C and the surface temperature were - 74"C, the vapor pressure at the dry surface would be approximately 100 times that at the drying boundary, 1 X 10 3 torr compared to 1 X torr. The temperature of water at the drying boundary, the fourth variable, is difficult to evaluate. It may be relatively constant, shifting only slightly higher each time the radiative heat from the copper block is increased and dropping back to its base value over each ensuing hour until the tissue is dry. The comparison of the drying data obtained with the partial pressure analyzer with the model developed for pure ice at constant temperature suggests that this is indeed the case, except during the first hour or two of drying, when the shell is so thin that it is not yet a very efficient insulator, Actually measuring the temperature of the core of frozen tissue is difficult. It is a simple matter to embed a fine thermocouple in a block of tissue and then freeze it, but an insulating shell quickly forms around the thermocouple as well as on the outer edge of the block, and the temperature of the block appears to rise even when the surrounding environmental temperature is kept constant. Stumpf and Roth ('67) measured drying rates with a n electrobalance, obtaining decreasing weight versus time, a measure equivalent to V = f(t) rather than dV1dt = f(t). The above hypothesis lends itself to a n explanation for why they found a n exponential decline in loss of water rather than a decline based on a cubic equation. Without a controllable source of external heat in their drier, the growing shell insulated the core of ice more and more effectively during the course of drying, and the temperature of the remaining frozen tissue dropped as a function of evaporative cooling. The vapor pressure of water at the dry surface remained the same, and that at the drying boundary dropped concomitantly with the drop in temperature. An exponential decline in the loss of water from the blocks would be expected under those circumstances. FREEZE-DRYING OF TISSUES FOR ELECTRON MICROSCOPY Apart from theoretical considerations, our technique is eminently practical, The unit shown in figure 2 can be constructed in any high quality glass shop and used according to the time schedule indicated, A partial pressure analyzer is unnecessary, A n even simpler drier can be constructed using Erlenmeyer vacuum filtering flasks for accommodating the specimen holder and molecular sieve, and a low temperature thermometer for monitoring the temperature of the copper block (Terracio and Coulter, '75). Tissue can be dried routinely in one day. Correctly executed, the technique is entirely reliable. Fixation and embedding Investigators have occasionally prepared frozen-dried tissue and examined it without fixation (Sjostrand and Kretzer, '75) hoping to get some image of the material in as natural a state as possible. The rationale for fixing the tissue in this work with Os04 vapor was to facilitate comparisons with ordinary methods of fixation in aqueous media, so that a critical evaluation of morphological preservation could be obtained. As laudable as the efforts have been to study unfixed tissue, it seemed a more pressing priority to evaluate the technique first with fixed tissue. There are two phases to the process of fixation. First, Os04vapor diffuses throughout the dry block, accumulating at all exposed interfaces and presumably dissolving in the lipid components and reacting with double bonds. The greyish cast of the tissue at this stage may occur as a result of those reactions. There is probably little chemical interaction with proteins, carbohydrates, and the polar ends of lipid molecules during this period. The second phase of fixation occurs during infiltration with resin, during which time the tissue turns completely black, taking on the typical appearance of 0smicated tissue, The presence of the liquid environment seems to be the key factor in permitting the reduction of the OsO,, because any epoxy resin, alcohol, or even paraffin causes the tissue to turn black. This is not due merely to a change in refractive index. If the resin is removed from the infiltrated blocks and they are redried, they remain black, although they have a porous appearance. The fact that extremely large blocks 49 1 turned black rather than grey during exposure to vapor is probably due to the fact that water remains in the cores of those blocks and thaws when the freezedrier is brought to atmosphere. Then with the liquid phase present, the OsO, goes into solution and the tissue is fixed in an aqueous medium, reducing the OsO, in a conventional manner. Then, as the block remains in the fixing chamber, it gradually becomes desiccated. Light microscopy Light microscopic examination of blocks is very helpful in finding tissue that is satisfactory for electron microscopy. The best blocks can be located, and the surface of the tissue that was initially frozen can be found. Then, the block can be oriented longitudinally and trimmed so that each thin section will be 1 mm or more in length and include 100 pm of good tissue plus several micrometers of excess plastic. The full length of the block can then be evaluated in the electron microscope. Often one segment has better tissue in it than the rest, and that part can be isolated for critical work. The densely stained and compressed tissue found in the central parts of the 1-2 mm blocks probably results from the same process that occurs when extremely large blocks turn black during vapor fixation. The central core of ice does not dry completely during initial freeze-drying, and violent surface tension effects occur subsequently during thawing and vapor fixation, Electron microscopy The most meaningful standard for evaluation of a technique of tissue preparation for electron microscopy is the reproducible preservation of fine structure. The reason that freeze-drying for electron microscopy has been so difficult is that useful tissue and reproducible results are quite difficult to obtain. Considered with the fact that a week or more of drying time has frequently been spent for a single experiment, it is not surprising that the electron microscopic literature of fkozen-dried tissue is very limited. By far the most troublesome problem with any freeze-drying technique is the formation of cytoplasmic reticulations caused by the growth of ice crystals. The 492 H . DAVID COULTER AND LOUIS TERRACIO physical aspects of ice crystal formation have been discussed in detail by Luyet (‘60) and will not be reviewed here. The tissue damage occurs when solutes and insoluble components of the cytoplasm are pushed to the periphery of the growing crystal. During lyophilization these components do not move back to their original sites in the cell but remain as a shell surrounding the area where the ice crystal grew. These shells of protoplasm are then fixed with OsO4 vapor with uncompromising clarity, and appear as reticulations of electron dense material scattered throughout the micrograph. They are absent only when the tissue was brought to a low temperature so quickly that water passes from a liquid to a frozen vitreous state. Such a field of tissue was shown at high resolution in figure 9. For intermediate magnifications and resolution, 1-10 nm reticulations are of no practical significance, and at even lower magnifications, 10-100 nm reticulations can be tolerated. The mitochondrial shrinkage spaces observed in frozen-dried tissue are puzzling. Their exclusive correspondence with the dense mitochondria may simply be an accident of freeze-drying, resulting from some minor biological variation, or they may be reflective of significant biological differences in the two mitochondrial populations. The dense mitochondria may have a higher content of electrolytes, or their membranes may have low permeabilities. These conditions could occur intermittently in relation to normal variations in functional state or as a result of hypoxia due to handling prior to freezing. In either case, an increased level of electrolytes would be expected to reduce the vapor pressure of the associated water, thereby slowing its removal. A reasonable hypothesis is that the dense mitochondria contain water whose outward diffusion is markedly hindered as the other water is removed. This remaining water becomes liquid when the drier is brought to atmosphere at room temperature, and the shrinkage results when the tissue is subsequently evacuated, shrinkage resembling in microcosm the distortions that occur when one tries to dry blocks that are too large. Frozen-dried Os04-fixed tissue has very high inherent contrast. This is not surprising, when the nature of the fixation process is considered. With no solvent interference, OsOI diffuses easily through the tissue and accumulates at every interface in the cell, dissolving in the lipids to the point of saturation, and coating the surface of every exposed protein and carbohydrate molecule. Probably not all of the G O 4 is reduced when the tissue is infiltrated, but i t is physically bound at every potentially reactive site, and is reduced immediately when the solvent becomes available. During aqueous fixation, on the other hand, the penetration of a01 is significantly limited, its accumulation on various structures requires chemical interactions, and many structures or molecules which might have reduced i t are lost by solvent extraction before the reactions can occur. The inherent contrast of membranes in this tissue is especially noteworthy. The O s 0 4 is taken up quite selectively by the inner lamina of the trilaminar membrane. The outer lamina is not usually visible unless the sections are stained with uranium and lead salts. Even then it is much thinner and less electron dense than the inner lamina (fig. 9). Successful application of the technique of freeze-drying for electron microscopy is of considerable importance to many problems in cell biology. Enzyme cytochemistry at the electron microscopic level, radioautography of water soluble tracers, and delineation of labile organellar and supramolecular relationships are but a few of the important potential applications. ACKNOWLEDGMENTS The authors thank Doctor William Sudderth for assisting with the mathematical derivations, and Doctors Morris Smithberg and Stanley Erlandsen for critically reviewing the manuscript. LITERATURE CITED Gersh, I. 1973 Submicroscopic Cytochemistry. Chap. 1. I. Gersh, ed. Academic Press, New York and London, pp. 1-15. Luyet, B. 1960 On various phase transitions occurring in aqueous solutions at low temperatures. Ann. N . Y. Acad. Sci., 85: 5 4 9 4 6 9 . Meryman, H . T. 1960 Principles of freeze-drying. Ann. N. Y. h a d . Sci., 85: 630-640. Rasmussen, D. H.,and A. P. MacKenzie 1971 The glass transition in amorphous water. A p FREEZE-DRYING OF TISSUES FOR ELECTRON MICROSCOPY plication of the measurements to problems arising in cryobiology. J. Phys. Chem., 75: 967973. Richards, A. G., Jr., T. F. Anderson and R. T. Hance 1942 A microtome sectioning technique for electron microscopy illustrated with sections of striated muscle. Proc. Soc. Exp. Biol. Med., 51 : 148-152. Richards, A. G., Jr., H. B. Steinbach and T. F. Anderson 1943 Electron microscopic studies of squid giant nerve axoplasm. J. Cell. Comp. Physiol., 21: 129-143. Sjostrand, F. S. 1943 Fixering och preparering for elektronmikroskopisk undersokning av vavnad. Nord. Med., 19: 1207-1212. Sjostrand, F. S., and F. Kretzer 1975 A new freeze-drying technique applied to the analysis of the molecular structure of mitochondria1 and chloroplast membranes. J. Ultrastruct. Re%, 5 3 : 1-28. Spurr, A. R. 1969 A low-viscosity epoxy resin embedding medium for electron microscopy. J. Ultrastruct. Res., 26: 31-43. Stephenson, J. L. 1953 Theory of the vacuum 493 drying of frozen tissues. Bull. Math. Biophys., 1 5 : 411429. Stirling, C. E., and W. B. Kinter 1967 Highresolution radioautography of galactose-aH accumulation in rings of hamster intestine. J. Cell Biol., 3 5 : 585-604. Stumpf, W. E., and L. J . Roth 1967 Freezedrying of small tissue samples and thin frozen sections below -6OOC. A simple method of cryosorption pumping. J. Histochem. Cytochem., 1 5 : 243-251. Terracio, L.,and H. D. Chulter 1975 Preparation of frozen-dried tissue for transmission electron microscopy: A simple and dependable method. 33rd Ann. Proc. Elect. Microscopy Soc. Amer., pp. 614415. Van Harreveld, A,, and J. Crowell 1964 Electron microscopy after rapid freezing on a metal surface and substitution fixation. Anat. Rec., 149: 3 a 1 3 a 6 . Van Harreveld, A., and S. K. Malhotra 1966 Demonstration of extracellular space by freezedrying i n the cerebellar molecular layer. J. Cell. Sci., 1 : 223-228.