Human cochlea was initially described by Eustachi in 1564 and first published by Albini in Leyden in 1744. Later, the Neapolitan anatomist Cotugno (1761) described the human inner ear and cochlear anatomy using corrosion casts in De aqueductibus Auris Humana Interna. Ninety years later, Corti (1851) discovered the receptor organ named after him. In a monumental work, Retzius (1884) improved histologic techniques and presented meticulous descriptions of the cellular anatomy of the human cochlea including surface preparations.
For fine-structure analyses of the human cochlea, good fixation is critical (Wright, 1980b; Gleeson, 1985; Osborne et al., 1989; Comis et al., 1990). In earlier studies, most human inner ears have been studied using perilymphatic fixation some hours after death (Bredberg, 1968; Bredberg et al., 1970). Several studies were performed on mammalian cochlea using SEM, but only a few on the human cochlea (Lim and Lane, 1969; Lim, 1972; Hunter-Duvar, 1975; Hoshino, 1977; Nomura and Kawabata, 1979; Wright, 1980b, 1981a, b, 1982, 1983; Nadol, 1983; Nadol and Burgess 1994; Bagger-Sjöbäck and Engström, 1985; Gleeson, 1985; Engström et al., 1987, 1990; Osborne et al., 1989; Comis et al., 1990; Hoshino, 1990; Reiss, 1990; Reiss and Vollrath, 1990; Pujol et al., 1991). Some were assessed in embryos (Tanaka et al., 1979; Igarashi, 1980; Fujimoto et al., 1981; Hoshino and Nakamura, 1985; Lavigne-Rebillard et al., 1985a, b; Lavigne-Rebillard and Pujol, 1986, 1987; Hoshino, 1990; Pujol et al., 1991). Here, we present results from ultrastructural analyses of human inner ear specimens obtained at surgery (Fig. 1). Swift fixation can be executed in the operating room to improve cell preservation. Bone minerals were removed with 10% Na-EDTA. Besides, it is possible to carry out immunohistochemistry with confocal microscopy for detailed protein detection. The surgical technique was described earlier (Tylstedt et al., 1997). Tissue can be regarded as “near-normal,” because the tumor extension does not involve the cochlea and audiometry displayed normal age-related tone thresholds. However, drilling could cause some damage, and structural/physiological changes might occur in middle-aged individuals despite normal hearing. Results show some discrepancies in anatomical structure and protein expression between rodents and man highlighting that men are not simply big mice.
A particularly important clinical issue is the slow retrograde degeneration of human auditory neurons following hair cell loss. This has immense significance and makes the cochlear implantation (CI) technique achievable over time. Experience suggests that functional neurons may remain; a blessing for the CI patient. The reason for this may reside in factors protecting the neurons such as neural architecture, neuron interaction, secreted protectants, or simply the inborn biological turn over or life span. However, these factors are important to explore. We do not yet recognize if electric stimulation may exert neurotrophic effects on human auditory neurons or influence their survival. Moreover, experimental models diverge here. Some studies in animals have demonstrated survival effects with electrical stimulation (Lousteau, 1987; Hartshorn et al., 1991; Leake et al., 1991, 1995, 1999; Miller et al., 1995; Mitchell et al., 1997; Kanzaki et al., 2002), while other studies report no rescue of SGNs associated with electrical stimulation (Shepherd et al., 1994, 2005; Araki et al., 1998; Li et al., 1999; Coco et al., 2006; Agterberg et al., 2010). A temporal bone study of 11 cochlear implant patients' temporal bones showed no evidence of enhanced neuron survival in the implanted ear compared to the unimplanted ear (Khan et al., 2005).
HUMAN COCHLEA—GROSS MORPHOLOGY
The two human cochleae are mirror-shaped, fluid-filled, coiled, fairly symmetrical bony tubes (3.2–4.2 cm long) situated in the petrous pyramids of the temporal bones. Perilymph; the fluid inside the scalae vestibuli and tympani communicates with the CSF via the cochlear aqueduct (Fig. 2). This route of fluid communication can allow spread of infections between these fluid compartments. The human cochlea is composed of ∼2 and three-fourth turns (Hardy, 1938), but unusual anatomy with cochlear three turns has been described (Tian et al., 2006). It is surrounded by a compact bony structure; the otic capsule. It is the hardest bone in the body with a trilamellar arrangement with islands of modified cartilage and high-mineral content, which increases the stiffness of the bony labyrinth (Bast, 1942). This is desirable, so that vibrations of fluid in the cochlea are reflected and not absorbed by the temporal bone. At dissection (Fig. 2), the otic capsule contrasts in yellow color to that of the surrounding cortical bone.
There is now much focus on cochlear anatomy due to surgical approaches to electrically stimulate the auditory nerve. Studies of plastic corrosion casts, made from unselected temporal bones harvested at autopsy (Wilbrand et al., 1974, Wadin, 1988), show the extensive anatomic variations in dimension, form, and shape (Dimopoulos and Muren, 1990; Erixon et al., 2009; Rask-Andersen et al., 2010) (Figs. 3–5). Also, the results from histologic reconstruction measuring the organ of Corti length demonstrate these variations (Hardy, 1938). Hardy found that the length of the Corti organ can vary as much as 10 mm. Interesting is the high degree of bilateral symmetry, something that characterizes paired organs in the body. Surgically, the basal end of the cochlea is of great interest. It curves in three dimensions (ant.-post./lateral-medial/inf.-sup.), resembling a “fish hook.” Here, a cochleostomy is made, and it is the site of the round window (RW). The hook anatomy varies which makes it difficult for the surgeon to optimally place the cochleostomy and reach scala tympani (ST) without destroying any inner ear structures.
Narrowing and remarkable coiling of the cochlea can occur that influences electrode array insertion even in normal cochleae, especially when using longer electrodes. The outer cochlear wall had a mean length of 42.0 mm (Table 1), while the first turn was 22.6 mm (range, 20.3–24.3 mm) representing 53% of the total length. Because the apical turn is small, full insertion of electrode array is hardly achievable, at least not with today's techniques. Such a deep insertion is hardly needed (and may not be advantageous), because the human spiral ganglion extends only 1 and three-fourth turns (i.e., 630°). The initiation site of the APs is not identified with certainty but may be the ganglion-cell soma in the modiolus as well as Ranvier nodes in the peripheral or central axons depending mainly on electrode position (Rattay et al., 2001a, b). Selective ion channels are now being analyzed at our laboratories.
Table 1. The total length of the outer wall excluding the basal half of the round window (RW)
New imaging techniques such as CT with 3D volume-rendered reconstructions can give thorough assessments of the coiling pattern; size and orientation of the cochlea relative to the skull base (Martinez-Monedero et al., 2011). This is of value before CI surgery especially in patients with malformations. Internal dimensions of the cochlear scalae and cochlear morphology related to CI were presented by Kawano et al. (1996), Verbist et al. (2009), and Biedron et al. (2010). The new technique called “cone-beam CT” improves even further image resolution and gives detailed information of the minor details of the temporal bone that is useful in CI.
Clinical Implications of Anatomic Variations
The large variations in cochlear lengths, angles between turns, and position in the skull base can influence the straightforwardness for the insertion of a CI electrode particularly passing the first turn. Most frequently damaged are the spiral ligament at the junction of the first and second half of the first turn, basilar membrane (BM), and osseous spiral lamina (Kennedy, 1987). Underdeveloped cochleae may show greater differences in the angle between the first and second turns and a smaller length of the base of the cochlea as evaluated from CT (Martinez-Monedero et al., 2011). A sharp bend of cochlear coiling in this region may hamper electrode insertion at this site. It increases the risk for perforations of cochlea membranes when the electrode tip moves from the first to the second turn. Interestingly, authors have also found that the cochlea may change its orientation during the first years of life that may require a modification of the opening of the facial recess (Lloyd et al., 2010; Martinez-Monedero et al., 2011).
HUMAN RW ANATOMY
“This is a plea of a practicing otologist rather than an experimental scientist for a re-evaluation of the round window. The writer feels that it is in this region that much future progress in intratympanic surgery will occur” (from Sellers, 1961)”
Today, there is more interest to save residual hearing when performing CI surgery, and the utility of a cochleostomy has now been questioned by several authors. Many surgeons therefore use the RW insertion technique again. This is possible through the design of new atraumatic electrode arrays. The round window membrane (RWM) is located at the end of the ST where the hook bends posteromedially. From the exterior, it is located under a bony overhang (tegmen) formed by the promontory wall of the cochlea. It is hardly seen by the surgeon due to the niche and mucosal fold. Initially, the technique used to insert a cochlear implant was through the RW. This was later abandoned for the drilling of a cochleostomy to enter the ST (Michelson, 1971; House and Urban, 1973; Clark, 1975, 1979; Burian et al., 1981). Because of the modiolus being in the way and the crista fenestra when using the RW approach, it was considered better to make a cochleostomy a few millimeters in front of the RWM (Clark et al. 1979) for the positioning of the electrode along the outer wall. Fig. 6a,b shows where surgeons may choose to drill a cochleostomy.
RWM was described by Scarpa (1747–1832) in 1772. He gave credit to Gabriele Fallopius (1523–1562) for discovering the RWM and Fabricius de Aquopendente (1538–1619) for the first pictorial delineation. In a more recent translation edited by Sellers and Anson (1962), young Scarpa's thorough description can be accessed and appreciated. They already then disagreed on the true shape of the RWM. Scarpa states that this membrane is not flat as anatomists had thought but instead conical. Scarpa proposed the term secondary tympanic membrane and believed that it played an even larger role for the transmission of sound into the inner ear. Adam Politzer referred to the RW membrane as the RW of Scarpa. Bast and Anson (1953) and Bollobas (1972) described its embryology in great detail.
Dimensions of the Human RW Membrane
The membrane itself has an average size of 2.3 mm × 1.87 mm according to Proctor et al. (1986) (Table 2). It does not grow postnatal. Su et al. (1982) found the minimal diameter to be 1.08 and the maximal diameter 2.28 mm based on data collected on as much as 463 specimens. Many authors do not seem to have considered the ovoid shape of the RW membrane. This can be due to the histological studies where it is difficult to realize its three-dimensional structure. The RW in human is not round but ovoid with a long and a short diameter. Adult dimensions are reached early during fetal development. Although individual values vary considerably, placement of a standard CI array with a maximum diameter of 1.0 mm through the RWM should be possible in most cases. Several electrodes today have dimensions below this size of 1 mm. RWM is formed like a saddle or “pringle-chips” (Alec Salt, oto.wustl.edu/cochlea/cachad). It consists of an infero-anterior vertical (4/7) part and a supero-posterior horizontal part (Okuno and Sando, 1988). In the posterior part, the membrane lies very close to the lamina spiralis ossea with a distance of around 0.1 mm (Figs. 2, 5, 6). In the center, this distance is around 1 mm (Franz et al., 1987). The thickness of the membrane varies between 50 and 100 μm and separates the fluid of the internal ear from the gas-filled space of the middle ear. The thinnest part of the RWM is the center, which can rupture if ST perilymph hydrostatic pressure rises relative to middle ear pressure. The middle ear mucosa that covers the outer surface of the RWM contains blood vessels and nerve fibers.
The RW membrane is suspended in the otic bone through a fibrous ring. The epithelium facing the middle ear is of low columnar or squamous type and consists of cells held together with tight junctions and a basal lamina. There is a thin lamina propria that contains connective tissue, nerve fibers, and blood vessels (Fig. 7). A secondary membrane (pseudo-membrane) or fold usually tapers the RW niche but not always. It can be incomplete and adherent to the RWM. One must therefore be careful not to remove this membrane directly during surgery, because it can tear the RWM. A small nerve innervating the RWM from the spiral ganglion region penetrates the fibrous layer. Theoretically, and because there are morphological similarities with other mechanoreceptors in the body, it could sense and monitor perilymph pressure variations (Rask-Andersen et al., 1999; Rask-Andersen and Illing, 2004). Interestingly, these anatomical structures were not found in quadrapods, suggesting that up-right posture in humans may require additional regulatory mechanisms for fluid homeostasis (unpublished data-Rask-Andersen).
The entrance to the RWM forms a small space called the fossula fenestrae rotunda. Its maximal diameter is 2.98 ± 0.23 mm (Takahashi et al., 1989). Its shape varies greatly (Wysocki, 1998; Toth et al., 2006), and its appearance may puzzle the surgeon and make him believe that the RWM has been erroneously misplaced with a partly ossified entrance mistaken for the membrane itself.
Anatomy of the Facial Recess and RWM Surgery
The space between the lateral aspect of the third portion of the facial nerve and the chorda tympani is called the facial recess and is the most common pathway to the middle ear to place a CI (Fig. 8). Its size and shape can vary greatly. Sometimes, the facial nerve can protrude and make visualization of the round window membrane (RWM) impossible without removal of some part of the posterior bony wall of the meatus. The illustrations below show the various examples of the visual exposure of the RWM from the facial recess approach according to Hamamoto et al. (2000) (Fig. 9).
ELECTRON MICROSCOPY OF THE HUMAN COCHLEA—RELEVANCE FOR CI
Studies of the fine structure of the human cochlea may give us better understanding of the intracochlear tissue interacting with an electrode during insertion, at place, and during electric stimulation. Here, we used the technique of field emission scanning electron microscopy of human cochleae. SEM of hemisectioned human cochlea is shown in Fig. 10. After decalcification, the cochlea was divided mid-modiolarly using a razor blade. It shows the typical cavities of the ST, scala vestibule (SV), and scala media (SM), or cochlear duct (Fig. 11). Variations in their shape and size are common. The pear-shaped ST at the second turn is notable, because the walls may interfere with a deep electrode insertion. Reissners membrane is very gracile and consists of both a mesothelial layer facing SV and an epithelial side facing SM (Fig. 11). The small organ of Corti rests on the BM. Some trauma was induced by the razor cut of the decalcified cochlea.
Organization of the Human Cochlea
Located within the organ of Corti (Fig. 1) resides the mechanosensory cells arranged in one row of inner hair cells (IHCs) and three to four (sometimes as many as five) rows of outer hair cells (OHCs). Bundles of sensory hairs (stereocilia) emerge from the apical poles of the sensory receptor cells. Hair cells are surrounded by several types of supporting cells, and these support cells have contact with the BM. Because hair-cell stereocilia are coupled to the overlying acellular tectorial membrane, oscillations of the BM cause back-and-forth deflection of the hair bundles. This motion is not uniform along the cochlear spiral, because the BM is narrow and rather stiff on its base but is wider and more compliant near the apex of the cochlear duct. von Bekesy (1960) was the first to show that sound sets up a travelling wave and its peak amplitude varying with respect to location on the BM with frequency. High frequencies are mapped near the base while low frequencies cause maximal vibrations closer to the apex. This leads to a tonotopical organization of the sensory cells where the location of excited hair cells along the BM determines the perceived frequencies. BM and the thin Reissner's membrane (RM) subdivide the spiral canal into its three compartments. SM is filled with endolymph that can be distinguished by its ion composition that is much like an intracellular character with respect to its potassium content. ST and SV are filled with perilymph containing an extracellular-like ion composition and communicated at the apex of the cochlea via the helicotrema. The fluid contained within the tunnel of Corti and Nuel's spaces within the sensory epithelium resembles perilymph's composition.
The human Reissner's membrane (RM) consists of an epithelial-cell layer facing the endolymph compartment of the SM and a mesothelial facing the perilymph compartment of the scala vestibuli (Fig. 10). These cell layers are separated by a basement membrane and form an important barrier between these two fluids that possesses completely different ion compositions. RM is involved in homoeostasis and fluid transport. Integrity of this membrane is essential for hearing to maintain the endocochlear potential (EP; +80 mV) in the cochlear duct. In man, the membrane differs markedly from that in rodents and monkeys in that the epithelial cells have an irregular shape, and their borders make a jigsaw pattern (de Fraissinette et al., 1993) with clusters of protruding cells. Infants and children share the same uniformly polygonal surface structure of the epithelial layer (Felix et al., 1993). The function of the discontinuous mesothelial cell layer and melanocytes located almost exclusively on the perilymphatic side remain unclear. A scala vestibuli electrode insertion would most probably disrupt the patients RM and abolish the EP, thereby losing any residual hearing that may have been present.
BM and Tympanic Covering Layer
The specialized basement membrane of the sensory epithelium is a trampoline-like array of elastic fibers suspended between the modiolus and the lateral wall. It consists of mainly radial running fibrils with a characteristic rectangular shape (140 A in diameter; Kimura, 1984) embedded within a matrix. These fibers contain several different kinds of collagen (Thalmann, 1993; Cosgrove et al., 1996a, b; Dreiling et al., 2002). Furthermore, proteoglycans (Tsuprun and Santi, 2001) and fibronectins (Cosgrove and Rodgers, 1997) have been identified within the matrix.
Mechanical properties along its spiral way vary considerably and enable the tonotopical organization of the basilar membrane (BM). At the base, this membrane is narrow and thick (0.1 mm), and it gets thinner and broadens toward the apex (0.5 mm). Therefore, the risk of perforation of the BM with an electrode array increases with depth of insertion. The portion of the BM between the tunnel of Corti and the osseous spiral lamina forms a continuous membrane with small apertures for the passage of nerve fibers to the organ of Corti, that is, the “habenulae perforatae,” while parallel strands that are peripheral to this area are free to vibrate. Along the ST, a cellular layer of tympanic covering cells that posseses phagocytic properties adheres to the thick matrix of the BM. It is possible that these cells react with the electrode by inducing a foreign body reaction leading to the formation of a fibroblast cover encasing the CI electrode. A robust proliferation of these encasing fibroblasts could lead to increased electrode impedances that can result in lower performance.
Designation as a ligament is an awkward denomination for this highly metabolically active and important tissue. The spiral ligament anchors the BM at the lateral aspect of the otic capsule. This attachment to the organ of Corti is characterized through the presence of tension fibroblasts that contain actin, myosin, and tropomyosin. This may generate a certain active tension that can be modulated (Henson and Henson, 1988). Besides its mechanical function, the spiral ligament plays an important role for the supply and drainage of perilymph. The spongiform appearance pervaded with an extensive capillary network suggests intense high level of communication between the ST and scala vestibuli. Therefore, the spiral ligament can be assigned to the domain of the perilymphatic compartment. The importance of this tissue for maintaining the ion balance is supported through the presence of gap junctions (connexins) and Na+/K+-ATPase pumps. The spiral ligament is thought to pump K+ out of the perilymph and transport it for maintaining the high concentration of K+ in the endolymph (Spicer and Schulte, 1991; Raphael and Altschuler, 2003). High-resolution SEM of mid-modiolar section of the human cochlea displays the interior surface anatomy of the lateral wall of the ST (Fig. 11a,b). This is where the tip of the electrode first reaches the spiral ligament at the junction between the lower and upper basal turn. This is a critical step during the CI insertion with risks for perforation of this very important structure, that is, the spiral ligament.
The human stria vascularis is composed of three distinct cell types in consecutive layers: marginal cells, intermediate cells, and basal cells (Fig. 11e,f). This high-metabolic tissue is served by an extensive meshwork of capillaries and forms the lateral aspect of the SM between RM and the spiral prominence. Marginal cells are derived from epithelial cells, face the endolymph lumen, and send long intermitting processes as far as the basal cells. The surface facing the endolymph lumen is exclusively formed by marginal cells and confined through tight junctions, numerous desmosomes, and varying density and size of microvilli, which border on the endolymphatic compartment of the SM (Kimura and Schuknecht, 1970; Wright, 1980a). The mesoderm-derived intermediate cells are melanocyte-like (Hilding and Ginzberg, 1977) and originate from the neural crest. A significant difference to other epithelia is the partial absence of a basement membrane underneath marginal cells. This may facilitate a close contact to the endothelial cells of strial capillaries and adjacent intermediate cells. The stria vascularis plays an essential role for generation and maintenance of the SM's EP. High expression of Na/K-ATPase, ionic pumps, and transporters and extensive vascularization highlights its energy consuming task. The lateral wall fibrocytes contain an amazingly large number of gap junctions. They form direct intercellular pathways for small molecules, ions, and electric currents. Mutational changes of Cx26 and also Cx30 have noted to be associated with congenital type of sensorineural deafness (Kelsell et al., 1997; Cohn and Kelley, 1999). These proteins are believed to be involved in the recirculation of potassium ions from the fluid and cellular spaces surrounding the hair cells into the supporting cells distal to the organ of Corti and shunting these ions into the fibrocytes in the spiral ligament. On freshly obtained human cochleas, we have identified several types of connexins in the organ of Corti and lateral wall tissues (Liu et al., 2009). Cx26 and Cx30 were found to be mostly coexpressed, suggesting that connexons in man consist of both types of these connexin proteins forming the channel (Fig. 11d). The exact molecular build-up in the human cochlea remains obscure. Connexin 26 was analyzed in human fetal cochlea by Kammen-Jolly et al. (2001). Connexin 43 mutations have been found to be associated with non-syndromal, autosomal, recessive deafness (Liu et al., 2001).
Tectorial membrane (TM) is an extracellular matrix that causes a shearing motion to stereocilia bundles when vibration enters the cochlea partition (Fig. 12). This auxiliary mass is required to present movement to the “light weight” and therefore rather vibration-insensitive hair cells. The TM is composed of radially running unbranched fibrils of type II and type IX collagen (type A) and highly branched fibers of type V collagen (type B) in which the thick fibers are embedded (Slepecky et al., 1992a, b, c). The jelly-like matrix is composed of various glycoproteins, for example, the tectorins and otogelin. Mutations in human α-tectorin underlie two dominantly inherited non-syndromic deafnesses, that is, DFNA8 and DFNA12 (Verhoeven et al., 1998). Stereocilia imprints from the OHC tallest stereocilia tips indicate a rather close attachment of these structures to the TM, while this close association between IHC stereocilia tips and the TM remains unclear. Stereocilia imprints as far as the distal end of marginal pillars suggest a different excitation for the last row of OHCs (Fig. 12b; Glueckert et al., 2005). Imprints of the IHCs and stereocilia tips on the human TM were described by Hoshino (1981). In our SEM studies, we did not observe any signs of imprints on the TM area of Hensen's stripe where the tips of the IHCs stereocilia seem to attach. There is no evidence that TM forms after birth in the human cochlea.15
Supporting cells provide tight and buckling resistant connections of hair cells with the BM. The most rugged of these support cells are the inner and outer pillar cells that contain a massive array of specialized tubular filaments (tonofibrils) that form the tunnel of Corti. Deiter cells sustain the bases of OHCs forming a “seat” for the basal portions of these sensory cells and send a phalangeal process to the reticular lamina completing their structural support. Hensen, Claudius, and Boettcher cells shape the lateral segment of the organ of Corti. Undulated lateral-cell membranes of Hensen and Claudius cells provide surface enlargement (Glueckert et al., 2005). Supporting cells provide more than only a mechanical function. Ion transporters and a well-expressed gap-junction system serve to maintain homoeostasis and enable the support cells to act in concert. Connexins play an important role to interconnect this organized system of support cells (Fig. 13). Gene products for several ion channels such as Kcc4 (Boettger et al., 2002) have been localized to supporting cells. Potassium that enters hair cells during transduction needs to be transported back to the endolymph to maintain the high-electrochemical gradient between endo- and perilymph. K+ channels and gap junctions are therefore needed to recycle potassium via hair cells, supporting cells, fibrocytes, and stria vascularis to the SM. Innervation of Deiters' and Hensen's cells has been described in the organ of Corti of several mammalian species (Wright and Preston, 1976; Fechner et al., 1998, 2001).
Hair Cells and Innervation
Two types of mechanotransductory sensory cells are present in the organ of Corti typically arranged in a single row of inner (IHCs) and three to four (sometimes even 5 rows of irregularly arranged OHCs in the apical region) rows of OHCs (Fig. 14). The distribution of ∼3,400 IHCs and 12,000 OHCs is less regular than in lower mammals with seemingly missing OHCs and supernumerary IHCs (Wright et al., 1987; Glueckert et al., 2005). Correlative studies of hair cell density and cochlear length were performed by Ulehlova et al. in 1987. A great variability in the length of the cochlear duct was found ranging from 28.0 to 40.1 mm. Average densities of the inner and outer hair cells decreased with increasing cochlear length with long cochleae having a greater number of hair cells in a given frequency region. These polarized neuroepithelial cells reveal specialized microvilli at their apical pole called stereocilia. These cudgel-shaped rigid (greek: stereos) cell protrusions contain a high amount of parallel-striated actin and are therefore unable to move like “real cilia.” This rigidity of the stereocilia decreases from base to apex and may be associated with declining resonance frequencies. Several proteins have been identified in animals that play an important role in interconnecting and maintaining the actin scaffold of the stereocilia (Flock et al., 1986; Slepecky and Ulfendahl, 1992). Mutations in the gene coding for the unconventional myosin VIIa that is only expressed in stereocilia cause deafness in humans. Stereocilia are anchored into the apical pole of the hair cells with electron-dense actin rootlets in the cuticular plate. This organelle, that is, cuticular plate is a thick terminal web with high content of actin and associated proteins, with the exception of the region of the basal body, which is called cuticular pore. Stereocilia are cross-linked among each other through side links that enables them to deviate in the same direction in concert. At their apical end, thin-tip links connect stereocilia of adjacent rows that are thought to be part of the mechanotransduction receptor channel. Apart from this common organization of the receptor pole, there are significant differences between the two types of auditory hair cells (Fig. 16).
Inner hair cells.
A single inner hair cell (IHC) posseses approximately 50-70 stereocilia at the basal end and 100 in the apex of the cochlea partition (Glueckert et al., 2005; Wright, 1984) IHCs are the primary afferent sensory cells of hearing. Each IHC is innervated by several afferents that are assumed to originate exclusively from large (type I) neurons in the spiral ganglion (Fig. 16). More than 90% of nerve fibers in the osseous spiral lamina are to be attributed to the large spiral ganglion cells (SGCs) reflecting the importance of IHC for afferent neurotransmission to the central auditory pathway. There is a maximum of 15 nerve fibers per IHC when compared with ∼26 nerve fibers/IHC in cats (Liberman, 1980a, b). The overall afferent innervation density is maximally 1,400 fibers/millimeter and that is very low compared to 3,000 fibers/millimeter in the cat. Each terminal has several synaptic contacts with one single IHC. As has been demonstrated in animals, these terminals in man are very susceptible to anoxia. They can react with enormous swelling as observed in postmortem material. Considerable variation in size of afferent terminals was also found in man (Nadol, 1990a). Large nerve endings frequently form multiple areas of synaptic specializations, branched, and synapsed with up to three adjacent IHCs. Electron dense specializations at the presynaptic side referred to as synaptic bars, bodies, or ribbons are a common feature in human cochlea specimens. Efferent nerve terminals only form synaptic contacts with afferent dendrites, like is the case for all mammals. Very few efferent terminals reach the cell body of the IHC.
Outer hair cells.
Very few afferent nerve fibers or basilar nerve fibers run at the basis of the tunnel to the outer hair cells (OHCs). Below the OHCs, they form the outer spiral fibers. They form huge bundles of spirally oriented nerve fibers of a group of several hundred between Deiter cells. They contain 0.1-μm-thick neurocanaliculi. A few efferent nerve fibers run through these bundles and can be easily distinguished through their larger size, neurofilaments, and “en passant” synapse with outer spiral nerve fibers.
The number of outer spiral nerve fibers increases in the upper turn to about 200 and 400 in the second turn, again reduced to 300 in the apical turn. At the base of OHCs, there are large and small afferent nerve endings. Possibly, the extensive system of outer spiral fibers may interconnect or synchronize groups of OHCs enhancing the organ of Corti responses of auditory stimuli. Another difference with regard to the animal cochlea is the fewer inner spiral efferent nerve fibers and efferent nerve endings at the base of the OHCs.
Immunohistochemical investigations of the human inner ear were performed in the last decade using antibodies against several different neurotransmitter proteins. GABA and acetylcholine have been demonstrated as efferent neurotransmitters of the human inner ear (Schrott-Fischer et al., 1994, 2002a). Glutamate-like immunostaining has also been detected in the human cochlea (Schrott-Fischer et al., 1994). The efferent cochlear nerve bundle and fiber population of the vestibulocochlear anastomosis were described in man by Gacek in 1961 and Arnesen in 1984.
Sensory Transduction and Cochlear Amplification
Hair cells are secondary sensory cells and therefore unable to generate an action potential. They communicate their responses to the central nervous system via their afferent nerve contacts to the type I SGCs in the cochlea that are able to generate spike potentials. These SGCs are located in the modiolus, most of them within Rosenthal's canal (basal turn), and send their peripheral processes via the osseous spiral lamina to the hair cells within Corti's organ. Although less numerous, the IHCs are the main sensory cells, providing synaptic input to 90–95% of the afferent nerve fibers of the major big type I SGCs. Small type II ganglion cells are also afferent but comprise only 5–10% of all spiral ganglion neurons and are thought to innervate only the OHCs. (Spoendlin, 1979, 1985; Spoendlin and Schrott, 1989) (Fig. 14) The motor protein prestin (Zheng et al., 2000; Dallos et al., 2006, 2008) enables OHCs to rapidly change both the length and the stiffness of their cell body as their main response to sound. This motile reaction may amplify the movement of the organ of Corti, increasing sensitivity and thereby contributing to frequency analysis and binaural acoustic localization. Immunohistochemical demonstration and localization of prestin in human OHCs were recently reported (Liu et al., 2009; Fig. 17a–c).
HUMAN SPIRAL GANGLION—ANATOMY
The human cochlea from normal hearing individuals contains ∼35,000 afferent neurons that are bipolar with soma (SGCs) located in the helical Rosenthal's canal (RC) in the modiolus along the 1¾ turns of the cochlea. The spiral ganglion terminates in a bulge containing the cell bodies of neurons innervating hair cells of the third turn (Ariyasu et al., 1989; Pamulova et al., 2006). Peripheral processes (named axons or dendrites) run within a bilamellar bony space of the osseous spiral lamina to the habenula perforata to exit this bony canal and innervate the hair cells (Fig. 18) (Rask-Andersen et al., 2006). At least two types of SGCs can be distinguished. Ninety percent can be assigned to large neurons (or type I; ∼20–30 μm in diameter). They are characterized by prominent nucleoli with loose chromatin and a cytoplasm rich in mitochondria and a well-developed endoplasmic reticulum (Rask-Andersen et al., 1997). Five to ten percent of SGCs are smaller neurons (or type II cells; ∼10–15 μm in diameter) with a nucleus that has less-pronounced heterochromatin, and the cytoplasm contains fewer mitochondria and neurofilaments. Morphological evidence of even three types of SGCs in the human was demonstrated (Nadol, 1990b; Rosbe et al., 1996). Small neuronal cells are believed to innervate the OHCs, while the large neuronal cells innervate the IHCs (Spoendlin, 1978, Spoendlin and Schrott, 1988, 1989). Efferent nerve supply is provided from the medial and lateral olivocochlear system with fibers running spirally within RC as the so-called intraganglionic spiral bundle. It contains smaller myelinated and unmyelinated efferent nerve fibers and are closely associated with smaller (type II) neuronal cells in the peripheral portion of RC. There are also adrenergic fibers running along blood vessels and some independent of blood vessels that are intimately related to the afferent nerve fibers. The smaller spiral ganglion neurons express the protein peripherin (Liu et al., 2010).
A characteristic feature of human SGC somata is their lack of a compact myelin layer around their perikarya (Fig. 19). Some myelination can be observed more frequently in cochlea specimens from elderly people than in specimens obtained from neonates (Arnold, 1987a, b). Ota and Kimura (1980) found that both large and small neurons were myelinated and unmyelinated. However, a majority of the population was unmyelinated (94%). According to Rattay et al. (2001a, b), this would indicate a slower electric conduction rate over the SGN region in man, and, theoretically, it would take a spike more than 0.3 ms to pass the unmyelinated soma and to arrive at the central terminating process. But, as recently discussed, this could be advantageous for generation of synchronous activity important for generation of complex speech signals allowing cross-excitation (Tylstedt and Rask-Andersen, 2001). In humans, SGCs form clusters or groups (Spoendlin and Schrott, 1990; Tylstedt and Rask-Andersen, 2001a) of neuronal cells that may embody innervation of tonotopically distributed hair cells. Each IHC receives synaptic contacts from 10 to 14 neurons (Spoendlin and Schrott, 1988). These groups of neurons may act in concert in the processing of electrical signals from corresponding hair cells forming “functional units.” Ganglion cells are not infrequently surrounded by common and incomplete satellite cells that could make electric cross excitation possible. Neural interactions were also found between nerve fibers and large cells as well as axosomatic connections between efferent and small ganglion cells (Rask-Andersen et al., 2000; Tylstedt and Rask-Andersen, 2001b). Nerve fiber synapses on SGCs in the human cochlea were observed by Kimura et al. (1979). Axosomatic synapses were observed on type II SGC in human (Arnold et al., 1980; Arnold, 1982a, b) and monkey (Kimura et al., 1987) but have rarely been found in lower vertebrates (Ivanov et al., 1992). Although the physiological significance of these synapses is unknown, they could mediate presynaptic neural modulation of type II auditory neurons at the level of the spiral ganglion.
HUMAN SPIRAL GANGLION AND CI
Electric stimulation from cochlear implants evokes APs in remaining acoustic nerve fibers. Where these spikes are initiated anatomically is not known. Nerve somata or initial axonal Ranviers nodes seem probable, because preserved dendrites may not seem to influence performance as evaluated histologically in patients treated with CI. Likewise, the amount of neurons necessary is still under debate (Linthicum and Galey, 1983; Linthicum et al., 1991; Nadol et al., 2001; Khan et al., 2005; Fayad and Linthicum, 2006; Nadol and Eddington, 2006), but astonishingly few neurons (∼10%) seem to be required to create speech performance, reflecting the redundancy present in the acoustic system. Thus, technical advancements seem to be needed in order to take full advantage of the potential of patients' residual neural populations. This should include novel stimulation principles, additional functional channels for enhanced frequency resolution, fine structure, and music perception. A problem is the anatomic/tonotopic compression and mismatch of the spiral ganglion neurons. Only in the basal turn are the ganglion cells distributed in such a way that frequency-specific stimulation seems possible with today's techniques (Greenwood, 1990, 1991). More apically, the ganglion cells reach only three-fourth of a turn, while the hair cell region extends to 1 and three-fourth turns over a rather short distance. These neurons, even though tonotopically arranged, cannot be selectively stimulated using current CI technique. Thus, CI stimulation or electrode insertion beyond 360° and 450° at present does not seem necessary. Also, modern technique uses monopolar stimulation that confines selective activation of neurons further questioning the need for deeper insertion of the CI electrode array. Because few data are available on the current spread in the highly anisotropic inner ear, predictions on directed stimulation of apical SGNs with a CI are difficult.
These questions are relevant to discuss in connection with new hearing-preservation techniques aiming at providing electric stimulation and hearing in the mid- and high-frequency regions. Electroacoustic stimulation technique uses both low-frequency acoustic hearing via the auditory hair cells and electric stimulation of preserved neurons in the same ear. Some patients may lose their residual hearing and become fully dependent on their implant for electrical stimulation of all frequencies. A deep insertion of the CI array increases the risk for loosing residual hearing but may theoretically improve performance in the case where a patient's residual hearing is eventually lost. Still, there is lacking information as to the optimal insertion depth in most cases with or without aims to preserve a patient's residual hearing.
Sensory-cell loss in animals leads to a retrograde Wallerian degeneration with a fall in the population density of the spiral ganglion's auditory neurons (Spoendlin, 1971; Schuknecht, 1993). This is believed to be caused by a disruption of neurotrophin secretion from the organ of Corti, which may act as survival factors for the associated auditory neurons. In patients with congenital deafness and adult-onset SNHL, neurons seem to be conserved even after many years of deafness. This observation suggests that there is a slower deterioration of SGCs in humans (Fayad and Linthicum, 2006). BDNF and NT-3 may be secreted by surrounding Schwann cells or satellite cells, because their receptors have been localized in the human spiral ganglion (Liu et al., 2011a, b). Also, receptors (c-ret for GDNF and isoforms) have been localized in the human spiral ganglion (Liu et al., 2011a, b; unpublished observations). It is highly plausible that several unrecognized intercellular signaling systems prevail in the human cochlea. Currently, we are focused on the surrounding satellite cell system, the unmyelination status of ganglion soma, cross-excitation, and connecting proteins, and their roles for the function, preservation, and survival of the human SGC.
Architecture of the Human Cochlea and Electrode Design—Today and the Future
An important issue is the design of various electrode arrays and how they adapt to the anatomical features of the human cochlea. Some electrode arrays are arranged close to the modiolar wall; so-called perimodiolar electrodes, while others lie free in the ST, mostly near the lateral wall (Fig. 20). Although the perimodiolar shape may result in a closer relationship between electrodes and SGC soma, it may also potentially cause trauma to the thin bony wall of the modiolus (Rask-Andersen et al., 2006) (Fig. 18). On the other hand, a laterally positioned electrode array may cause strain to the lateral wall and BM. In revision surgery, more force may be required to extract a CI that is hugged around the modiolus. This could be a disadvantage in children who can be expected to undergo revision surgery several times during their life time. In the future, electrode designers need to shape electrodes in such a way that they minimize trauma during insertion. This is important, because trauma against the medial wall of the ST can damage the spiral ganglion, which is the prime target of electric stimulation by the electrodes of the implant. Smaller and more flexible arrays will be needed to adjust to the individual structural variations in the design of the human cochlea and for preserving the residual hearing of individual patients.
Harvesting of tissue: Human inner ear specimens, without any known hearing impairment, were obtained at surgery for the removal of life-threatening large petroclival meningiomas compressing the brainstem (Tylstedt et al., 1997). The study conforms to The Declaration of Helsinki and was approved by the medical ethical committee at the Uppsala University Hospital (no. C254/4, C45/7 2007). Patient's consent was obtained.
Casting of the human cochlea: We analyzed 73 of 324 archival, nonselected, adult, corrosion casts of human inner ears. Anatomic reference points were constructed from photographic reproductions taken at different angles, and various dimensions were assessed using planimetry. Anatomic variants with particular clinical/surgical interests were pinpointed (Erixon et al., 2009). The results may assist to foresee new strategies for electric stimulation and also to better understand today's technology.
The cochlea was washed in phosphate-buffered saline, pH 7.4, dehydrated in graded ethanol (70, 80, 90, 95, and 100%; 10 min each), critical point-dried, and attached to aluminium stubs. The specimens were coated in a BALTECH MED020 Coating System with gold–palladium to a nominal depth of 10–12 nm and viewed in a ZEISS DSM982 Gemini field emission electron microscope operating at 5 kV. Maximal resolution at this voltage was estimated to ∼2 nm. Digital photos were taken at 1280–1024 ppi resolution. Measurements were performed using the image analysis software Image Pro 126.96.36.199 (Media Cybernetics, MD).
Confocal, fluorescent, and bright field imaging: Sections were imaged using bright field, fluorescent, and confocal fluorescent techniques. Bright field and fluorescent images were obtained using an inverted fluorescent microscope (Nikon TE2000, Japan) equipped with fluorescence unit and a spot digital camera with three filters (for emission spectra at 358, 461, and 555 nm) and mounted on a computer system coupled to an image software system including image merging and fluorescent intensity analyzer. For confocal microscopy, we used a Nikon TE300 microscope equipped with laser imaging system using three different filters (Liu et al., 2010; Rask-Andersen et al., 2010). Antibody combinations: Various types of antibodies, dilutions, characteristics, and source were used singly or in combination (merge) for better identification of the anatomical structures. For identification of various nerve elements in the decalcified cochlea, we used beta III tubulin as a marker of type I spiral ganglion cells while peripherin was used for the identification of type II spiral ganglion cells (Liu et al., 2010). Furthermore, we used prestin antibodies for the identification of outer hair cells and parvalbumin as a marker for hair cells, thus also including the inner hairs that were proven positive in man (Rask-Andersen et al., 2009). Antibodies against prestin were used for immunohistochemistry mostly in double-labeling fashion. The following antibodies and combinations were used. Antiparvalbumin antibodies combined with antiprestin (C-20, sc-22694, 1:300, and 1:100; Chemicon); antiperipherin and antiprestin (1:100 and 1:50; Chemicon); antineurofilament 160 combined with antiprestin (1:100 and 1:100; Chemicon); anti-Tuj1 (beta-III tubulin) combined with antiprestin (1:300 and 1:100; Chemicon). Parvalbumin immunoreactivity.