h
Serial Reconstruction of Inner Hair Cell Afferent Innervation Using Semithick Sections JONATHAN H. SIEGEL Departments of Communication Sciences and Disorders and Neurobzology and Physiology, Northwestern UniuersLty, Euanston, Illinois 60208
KEY WORDS
3-D reconstruction, HVEM, Synaptic contacts, Cochlea
ABSTRACT The afferent innervation pattern of inner hair cells in the apex of the guinea pig cochlea was studied using serial reconstruction of semithick (0.25-pm) sections and high-voltage electron microscopy (HVEM). This thickness produced a good compromise between the ability to resolve details of the synaptic contacts between the hair cells and sensory neurons and the number of sections required to reconstruct the nerve terminals within the receptor organ. The use of a goniometer allowed the sections to be tilted to angles optimum for viewing either the synaptic membrane specializations or the presynaptic bodies. Reasonably good images of 0.25-pm sections could be obtained using a conventional 120-keV microscope, but the images produced by the HVEM were clearly superior. The sensory nerve terminals and hair cells were reconstructed using a microcomputer-based computer-aided-design system. Nerve terminals with complex shapes could be successfully rendered as surface models viewed as stereo pairs. The advantages and limitations of the techniques used are discussed.
INTRODUCTION though the use of semithick sections made the task The innervation patterns of cochlear hair cells of sev- much easier, it was still a formidable effort and not
eral mammalian species have been studied using se- without disadvantages. The project initially used the rial-section electron microscopy (Hashimoto and high-voltage (1,000-keV) electron microscope (HVEM) Kimura, 1987; Liberman, 1980a,b; Nadol 1983a,b; at the University of Wisconsin, Madison. Some followSmith and Sjostrand 1961b; Spoendlin 1966; Takasaka up microscopy of the same series was performed on a et al., 1983). Even among mammals, the number of conventional 120-keV microscope. Although the discrete synaptic complexes seen between the inner HVEM produced clearly superior images, the most hair cells and individual sensory neurons of the spiral troublesome problems with the images produced by ganglion (radial afferent neurons) appears to be species both microscopes resulted from the overlapping of dependent (Liberman, 1980a; Nadol, 1983a; Spoendlin, small structures within a section. This is, in turn, re1966). These studies required several hundred serial lated t o the density of cytoplasmic structures in the thin sections to reconstruct the nerve cell processes cells of interest. The inner hair cells and olivocochlear within the sensory epithelium. Such a task is immense, efferent neurons have the greatest cytoplasmic density since as many as 25 sensory neurons may contact a and the ability to detect presynaptic bodies in the hair single inner hair cell. It is not surprising that few stud- cells determined the maximum usable section thickies of this kind have been performed and, consequently, ness. A popular computer-aided-design (CAD) system that there is currently little appreciation for changes in running on a personal computer provided a flexible yet means of reconstructing the series and the pattern from base to apex of the cochlea, or between inexpensive it graphically. The methods used are depresenting species. In the present study, we sought to reduce the scribed and the strengths and weaknesses discussed in number of serial sections required to reconstruct the detail to aid those who may wish to conduct similar afferent terminations on inner hair cells of the guinea studies. pig cochlea by cutting thicker (0.25-pm semithick) secPreliminary descriptions of our reconstruction of 35 tions. The advantages and disadvantages of such radial afferents within the organ of Corti of the guinea simithick sections have recently been reviewed (Marko pig cochlea have been presented (Siegel, 1987a,b).Each et al., 1988; Rieder et al., 1985). radial afferent neuron typically runs unbranched after The hair cell provides a convenient marker for the entering the sensory epithelium t o terminate on one location of active zones in the presence of a dense preinner hair cell with a single synaptic contact. A similar synaptic body, first described by Smith and Sjostrand pattern has been reported for the cat (Spoendlin, 1966; (1961a), attached to the hair cell membrane at the acLiberman, 1980a). tive zone (Gulley and Reese, 1977; Saito and Hama, 1984). This feature turned out to be particularly important for the present study, since the density of the Received December 23, 1988; accepted in revised form February 5, 1989. hair cell cytoplasm would otherwise make the task of Address reprint requests to Dr. Jonathan H. Siegel, Northwestern University, identifying active zones extremely difficult. Even 2299 Sheridan Rd. Evanston, IL 60208.
0 1990 WILEY-LISS. INC
198
J.H. SIEGEL
MATERIALS AND METHODS The right cochlea of a young adult guinea pig free of middle ear infections was fixed in vivo by gently perfusing the perilymphatic spaces with a mixture of aldehydes (2.5% glutaraldehyde, 1.0% acrolein, 0.1% paraformaldehyde, 20 mM MgC1, buffered with 0.09 M sodium cacodylate to pH 7.4). This fixative preserves the ultrastructure of both hair cells and neurons of the organ of Corti and produces relatively little swelling of the afferent neural dendrites (Siege1 and Brownell, 1986). The animal was killed soon after the perfusion was started, reasoning that fixation would be faster in the absence of blood flow. After approximately 30 min, the temporal bone was removed from the skull, several openings were made in the bony capsule, and the cochlea was placed in a vial of fixative and stored a t 4°C for 12 hr. The cochlea was then rinsed for 45 min in three changes of buffer and post-fixed in 1%OsO, in the same buffer for 1 h r at 4°C. Following several rinses, the cochlea was then stained en bloc with aqueous 1%uranyl acetate for 30 min and rinsed again in distilled water. Block staining is particularly useful for studies of thick sections, since penetration problems sometimes encountered with postembedding section staining are eliminated. Our uranyl acetate staining procedures routinely produced some precipitates, which were highly specific for some radial afferents
TUNNEL OF CORTI
-
---.
Modiolar Sector
Fig. 1. Orientation diagram depicting digitized profiles of the three inner hair cells in the first (closest to the hair cell nucleus) section in the series. The azimuth angle used for specifying the viewing orientation is as defined in the text.
(i.e., see Figs. 3,6). Since the degree of precipitation varied considerably form cell to cell, this artifact proved useful for identifying individual cells. The cochlea was then dehydrated in acetone and embedded in
Fig. 2. Tubular network (arrows) with associated mitochondria (M) seen in the presynaptic cytoplasm of an inner hair cell. Section thickness: 0.25 pm; accelerating voltage: 120 keV.
RECONSTRUCTION OF INNER HAIR CELL INNERVATION
199
Fig. 3. Inner spiral (olivocochlear) efferent varicosity (El in contact with radial afferent neurons (R). One radial afferent was densely stained with uranyl acetate precipitate. Although additional precipitate can be seen in the intercellular spaces, the artifact is not present in other cells. Section thickness: 0.25 pm;accelerating voltage: 1,000 keV.
epon-araldite. It was then dissected into segments and re-embedded as a flat surface preparation (Bohne, 1972). A block from the boundary between the third and fourth turns was oriented so that the long axis of the inner hair cells was perpendicular t o the block face. Cutting sections in this orientation proved optimal for reconstructing the radial afferents, since they too run primarily perpendicular to the block face. The typically circular profiles of afferents in each section facilitated the alignment of adjacent sections. Even with relatively thick sections, the profile of a given afferent does not usually change substantially from section to section, making it easy to keep track of the numerous cells studied. This block orientation also offers the possibility of following cells which leave the field of serial micrographs by rephotographing the adjacent areas. A series of 120 sections was cut from the block starting from the bottom of the nuclei of the inner hair cells. At a thickness of 0.25 pm, this covered the entire 30pm course of the radial afferents within the organ of Corti. Electron microscopy was performed using the 1,000keV (AEI EM-7) microscope at the University of Wisconsin, Madison. Follow-up micrographs were taken using a conventional 120-keV microscope (JEOL 1OOCX). Both microscopes have goniometer stages, al-
lowing the sections to be tilted in the electron beam to take stereo micrographs. Electron micrographs were taken of three adjacent inner hair cells at a magnification of 4,000 and printed a t a final magnification of approximately 12,000. To establish alignment marks for the entire series, the micrographs were copied on an IBM Copier I11 photocopy machine with a contrast setting optimized to reproduce cell outlines but to avoid excessive background contrast. This copy machine was found to produce satisfactory images of continuous-tone micrographs and produced minimal distortion and magnification change (
Serial Reconstruction of Inner Hair Cell Afferent Innervation Using Semithick Sections JONATHAN H. SIEGEL Departments of Communication Sciences and Disorders and Neurobzology and Physiology, Northwestern UniuersLty, Euanston, Illinois 60208
KEY WORDS
3-D reconstruction, HVEM, Synaptic contacts, Cochlea
ABSTRACT The afferent innervation pattern of inner hair cells in the apex of the guinea pig cochlea was studied using serial reconstruction of semithick (0.25-pm) sections and high-voltage electron microscopy (HVEM). This thickness produced a good compromise between the ability to resolve details of the synaptic contacts between the hair cells and sensory neurons and the number of sections required to reconstruct the nerve terminals within the receptor organ. The use of a goniometer allowed the sections to be tilted to angles optimum for viewing either the synaptic membrane specializations or the presynaptic bodies. Reasonably good images of 0.25-pm sections could be obtained using a conventional 120-keV microscope, but the images produced by the HVEM were clearly superior. The sensory nerve terminals and hair cells were reconstructed using a microcomputer-based computer-aided-design system. Nerve terminals with complex shapes could be successfully rendered as surface models viewed as stereo pairs. The advantages and limitations of the techniques used are discussed.
INTRODUCTION though the use of semithick sections made the task The innervation patterns of cochlear hair cells of sev- much easier, it was still a formidable effort and not
eral mammalian species have been studied using se- without disadvantages. The project initially used the rial-section electron microscopy (Hashimoto and high-voltage (1,000-keV) electron microscope (HVEM) Kimura, 1987; Liberman, 1980a,b; Nadol 1983a,b; at the University of Wisconsin, Madison. Some followSmith and Sjostrand 1961b; Spoendlin 1966; Takasaka up microscopy of the same series was performed on a et al., 1983). Even among mammals, the number of conventional 120-keV microscope. Although the discrete synaptic complexes seen between the inner HVEM produced clearly superior images, the most hair cells and individual sensory neurons of the spiral troublesome problems with the images produced by ganglion (radial afferent neurons) appears to be species both microscopes resulted from the overlapping of dependent (Liberman, 1980a; Nadol, 1983a; Spoendlin, small structures within a section. This is, in turn, re1966). These studies required several hundred serial lated t o the density of cytoplasmic structures in the thin sections to reconstruct the nerve cell processes cells of interest. The inner hair cells and olivocochlear within the sensory epithelium. Such a task is immense, efferent neurons have the greatest cytoplasmic density since as many as 25 sensory neurons may contact a and the ability to detect presynaptic bodies in the hair single inner hair cell. It is not surprising that few stud- cells determined the maximum usable section thickies of this kind have been performed and, consequently, ness. A popular computer-aided-design (CAD) system that there is currently little appreciation for changes in running on a personal computer provided a flexible yet means of reconstructing the series and the pattern from base to apex of the cochlea, or between inexpensive it graphically. The methods used are depresenting species. In the present study, we sought to reduce the scribed and the strengths and weaknesses discussed in number of serial sections required to reconstruct the detail to aid those who may wish to conduct similar afferent terminations on inner hair cells of the guinea studies. pig cochlea by cutting thicker (0.25-pm semithick) secPreliminary descriptions of our reconstruction of 35 tions. The advantages and disadvantages of such radial afferents within the organ of Corti of the guinea simithick sections have recently been reviewed (Marko pig cochlea have been presented (Siegel, 1987a,b).Each et al., 1988; Rieder et al., 1985). radial afferent neuron typically runs unbranched after The hair cell provides a convenient marker for the entering the sensory epithelium t o terminate on one location of active zones in the presence of a dense preinner hair cell with a single synaptic contact. A similar synaptic body, first described by Smith and Sjostrand pattern has been reported for the cat (Spoendlin, 1966; (1961a), attached to the hair cell membrane at the acLiberman, 1980a). tive zone (Gulley and Reese, 1977; Saito and Hama, 1984). This feature turned out to be particularly important for the present study, since the density of the Received December 23, 1988; accepted in revised form February 5, 1989. hair cell cytoplasm would otherwise make the task of Address reprint requests to Dr. Jonathan H. Siegel, Northwestern University, identifying active zones extremely difficult. Even 2299 Sheridan Rd. Evanston, IL 60208.
0 1990 WILEY-LISS. INC
198
J.H. SIEGEL
MATERIALS AND METHODS The right cochlea of a young adult guinea pig free of middle ear infections was fixed in vivo by gently perfusing the perilymphatic spaces with a mixture of aldehydes (2.5% glutaraldehyde, 1.0% acrolein, 0.1% paraformaldehyde, 20 mM MgC1, buffered with 0.09 M sodium cacodylate to pH 7.4). This fixative preserves the ultrastructure of both hair cells and neurons of the organ of Corti and produces relatively little swelling of the afferent neural dendrites (Siege1 and Brownell, 1986). The animal was killed soon after the perfusion was started, reasoning that fixation would be faster in the absence of blood flow. After approximately 30 min, the temporal bone was removed from the skull, several openings were made in the bony capsule, and the cochlea was placed in a vial of fixative and stored a t 4°C for 12 hr. The cochlea was then rinsed for 45 min in three changes of buffer and post-fixed in 1%OsO, in the same buffer for 1 h r at 4°C. Following several rinses, the cochlea was then stained en bloc with aqueous 1%uranyl acetate for 30 min and rinsed again in distilled water. Block staining is particularly useful for studies of thick sections, since penetration problems sometimes encountered with postembedding section staining are eliminated. Our uranyl acetate staining procedures routinely produced some precipitates, which were highly specific for some radial afferents
TUNNEL OF CORTI
-
---.
Modiolar Sector
Fig. 1. Orientation diagram depicting digitized profiles of the three inner hair cells in the first (closest to the hair cell nucleus) section in the series. The azimuth angle used for specifying the viewing orientation is as defined in the text.
(i.e., see Figs. 3,6). Since the degree of precipitation varied considerably form cell to cell, this artifact proved useful for identifying individual cells. The cochlea was then dehydrated in acetone and embedded in
Fig. 2. Tubular network (arrows) with associated mitochondria (M) seen in the presynaptic cytoplasm of an inner hair cell. Section thickness: 0.25 pm; accelerating voltage: 120 keV.
RECONSTRUCTION OF INNER HAIR CELL INNERVATION
199
Fig. 3. Inner spiral (olivocochlear) efferent varicosity (El in contact with radial afferent neurons (R). One radial afferent was densely stained with uranyl acetate precipitate. Although additional precipitate can be seen in the intercellular spaces, the artifact is not present in other cells. Section thickness: 0.25 pm;accelerating voltage: 1,000 keV.
epon-araldite. It was then dissected into segments and re-embedded as a flat surface preparation (Bohne, 1972). A block from the boundary between the third and fourth turns was oriented so that the long axis of the inner hair cells was perpendicular t o the block face. Cutting sections in this orientation proved optimal for reconstructing the radial afferents, since they too run primarily perpendicular to the block face. The typically circular profiles of afferents in each section facilitated the alignment of adjacent sections. Even with relatively thick sections, the profile of a given afferent does not usually change substantially from section to section, making it easy to keep track of the numerous cells studied. This block orientation also offers the possibility of following cells which leave the field of serial micrographs by rephotographing the adjacent areas. A series of 120 sections was cut from the block starting from the bottom of the nuclei of the inner hair cells. At a thickness of 0.25 pm, this covered the entire 30pm course of the radial afferents within the organ of Corti. Electron microscopy was performed using the 1,000keV (AEI EM-7) microscope at the University of Wisconsin, Madison. Follow-up micrographs were taken using a conventional 120-keV microscope (JEOL 1OOCX). Both microscopes have goniometer stages, al-
lowing the sections to be tilted in the electron beam to take stereo micrographs. Electron micrographs were taken of three adjacent inner hair cells at a magnification of 4,000 and printed a t a final magnification of approximately 12,000. To establish alignment marks for the entire series, the micrographs were copied on an IBM Copier I11 photocopy machine with a contrast setting optimized to reproduce cell outlines but to avoid excessive background contrast. This copy machine was found to produce satisfactory images of continuous-tone micrographs and produced minimal distortion and magnification change (
