JN Email Content Delivery
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
 QUICK SEARCH:   [advanced]


     


J Neurophysiol 96: 2653-2669, 2006. First published August 9, 2006; doi:10.1152/jn.00565.2006
0022-3077/06 $8.00
This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow All Versions of this Article:
96/5/2653    most recent
00565.2006v1
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Rowe, M. H.
Right arrow Articles by Peterson, E. H.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Rowe, M. H.
Right arrow Articles by Peterson, E. H.

Autocorrelation Analysis of Hair Bundle Structure in the Utricle

M. H. Rowe and E. H. Peterson

Neuroscience Program and Department of Biological Sciences, Ohio University, Athens, Ohio

Submitted 27 May 2006; accepted in final form 2 August 2006


    ABSTRACT
 TOP
 ABSTRACT
 INTRODUCTION
 METHODS
 RESULTS
 DISCUSSION
 GRANTS
 ACKNOWLEDGMENTS
 REFERENCES
 
The ability of hair bundles to signal head movements and sounds depends significantly on their structure, but a quantitative picture of bundle structure has proved elusive. The problem is acute for vestibular organs because their hair bundles exhibit complex morphologies that vary with endorgan, hair cell type, and epithelial locus. Here we use autocorrelation analysis to quantify stereociliary arrays (the number, spacing, and distribution of stereocilia) on hair cells of the turtle utricle. Our first goal was to characterize zonal variation across the macula, from medial extrastriola, through striola, to lateral extrastriola. This is important because it may help explain zonal variation in response dynamics of utricular hair cells and afferents. We also use known differences in type I and II bundles to estimate array characteristics of these two hair cell types. Our second goal was to quantify variation in array orientation at single macular loci and use this to estimate directional tuning in utricular afferents. Our major findings are that, of the features measured, array width is the most distinctive feature of striolar bundles, and within the striola there are significant, negatively correlated gradients in stereocilia number and spacing that parallel gradients in bundle heights. Together with previous results on stereocilia number and bundle heights, our results support the hypothesis that striolar hair cells are specialized to signal high-frequency/acceleration head movements. Finally, there is substantial variation in bundle orientation at single macular loci that may help explain why utricular afferents respond to stimuli orthogonal to their preferred directions.


    INTRODUCTION
 TOP
 ABSTRACT
 INTRODUCTION
 METHODS
 RESULTS
 DISCUSSION
 GRANTS
 ACKNOWLEDGMENTS
 REFERENCES
 
One of the major, unresolved questions in hair cell research is the significance of ciliary bundle structure. All bundles are built on the same basic plan—a staircase pattern of stereocilia, with or without a kinocilium—but the realization of this plan is highly variable. There are two reasons for believing this variability has information content, i.e., it is telling us something significant about how hair bundles work. First, it is orderly. Hair bundle structure differs systematically between species, between endorgans, and between different regions of a single epithelium (reviews: Eatock and Lysakowski 2006Go; Hackney and Furness 1995Go; Lewis et al. 1985Go; Platt 1983Go; Saunders and Dear 1983Go; Smotherman and Narins 2000Go). Second, mechanical behavior depends on structure (Gordon 1978Go). So the strikingly diverse structures of hair bundles suggest they will differ in mechanical behavior and thus play different roles in signaling head movements.

The significance of bundle heterogeneity has proved difficult to analyze. Two major obstacles are the difficulty of making accurate measurements of such small, delicate structures and of acquiring large enough samples for quantitative comparisons. We are left with a rich, qualitative literature on differences in bundle structure and, with the possible exception of freestanding auditory hair bundles in alligator lizard (reviews: Aranyosi and Freeman 2004Go; Fettiplace and Fuchs 1999Go), little idea of what these differences mean to the behaving animal.

We are using computational and experimental approaches to study bundle structure and its functional significance in an otoconial organ, the utricle (Fontilla and Peterson 2000Go; Moravec and Peterson 2004Go; Nam et al. 2005Go, 2006Go; Rowe and Peterson 2004Go; Silber et al. 2004Go; Xue and Peterson 2006Go). As part of this effort, we developed new methods to quantify differences in bundle structure. Here we apply one of these methods, autocorrelation analysis of scanning micrographs (Rowe and Peterson 2004Go), to quantify several mechanically significant features of utricular bundles and characterize their spatial variation.

Spatial variation in bundle structure can arise from three sources. 1) Directional variation. Utricular bundles are organized into a series of radial transects that fan out from the medial margin of the macula (Fig. 1A, gray arrows) (Lindeman 1969Go). The activation axes (axes of maximum sensitivity) (Lowenstein and Wersäll 1959Go; Shotwell et al. 1981Go) of bundles along a single transect have similar orientations; but average bundle orientation in neighboring transects differs, so bundles in each transect are maximally sensitive to a different direction of head movement. 2) Hair cell type. All vertebrates have type II hair cells; amniotes (reptiles, birds, mammals) have a second hair cell type (type I) (Wersäll 1956Go). Spatial variation in bundle structure can arise if type I and II bundles differ in structure and also have different spatial distributions. 3) Zonal variation. Utricular bundles differ with position along a single radial transect. For example, there are differences between bundles in the striola (a crescent-shaped specialization in the macula and overlying otoconial membrane) and the extrastriola, even for bundles with parallel activation axes (e.g., Baird and Lowman 1978Go; Hillman 1976Go; Lapeyre et al. 1992Go; Lewis and Li 1975Go; Lim 1976Go; Platt and Popper 1981Go; Severinsen et al. 2003Go; Xue and Peterson 2006Go). This zonal variation is one major focus of the present study. It is a ubiquitous feature of vertebrate utricles; this suggests that it plays an important role in enabling these organs to detect and encode head movement. Furthermore, zonal variation in bundle structure covaries with utricular afferent properties such as discharge regularity and response dynamics (Baird and Lewis 1986Go; Goldberg et al. 1990Go; review in Lysakowski and Goldberg 2004Go); thus it may contribute to physiological diversity in afferents.


Figure 1
View larger version (59K):
[in this window]
[in a new window]
 
FIG. 1. Schematic diagram of the utricular macula in T. scripta illustrating 2 sources of systematic spatial variation in hair bundle structure (A) and random variation in bundle orientation at a single macular locus (B). A: gray arrows, average activation axes of utricular hair bundles. Zonal variation along a medial-to-lateral transect that spans all zones was quantified; the approximate location of this transect is indicated by the line of numbers. We also analyzed random variation in bundle orientation in small circular areas that approximate the collecting area of utricular afferents (A, a and b). At evenly spaced locations along the transect (A, gray numbers), we quantified variation in the orientation of activation axes (B, arrows), defined as the hexagon axis closest to the bundle's axis of bilateral symmetry (ABS). In 1 analysis (a), we collected ≤25 samples at each location. These samples formed 2 concentric rings around a central sample. In a, the central sample is at location 6 and only 4 more widely spaced samples from the inner ring are illustrated for clarity. We eliminated samples that impinged on the borders of the transect, resulting in a total of 319 samples for the 2 utricles. In a 2nd analysis (b), we investigated whether varying sample radius from 10 to 40 µm (U50) or 50 µm (U5) would significantly increase the range of bundle orientations. In the figure, only four size variants of 1 sample at 1 location (3) are illustrated for clarity. B: stereociliary arrays. Arrows , hexagon axis closest to each bundle's ABS (axis 1), i.e., the probable activation axis of the bundle. Numbers indicate the angle between axis 1 and a line parallel to the transect. The orientation of axis 1 in these 6 neighboring cells differs by 27° (33–6°). Because such local variations in axis 1 orientation show no systematic pattern (see also Fig. 7), we refer to this as random variation. Scale: 1 µm.

 
In addition to these three types of systematic spatial variation in hair bundle structure, utricular bundles also exhibit apparently random variation in structure at any one macular locus. One example is the spread of hexagon orientations (and thus activation axes) in neighboring bundles (Fig. 1B). Such random variation may help explain why afferents respond to movement directions perpendicular to their axes of maximal sensitivity (Dickman et al. 1991Go; Fernandez and Goldberg 1976Go; Si et al. 1997Go). Thus a second focus of this study was to quantify variation in hexagon orientation and estimate the effect this might have on directional tuning in afferents.

To quantify zonal and random variation in utricular hair bundles, we used autocorrelation analysis of stereociliary arrays on the hair cell's apical surface (Fig. 1B). We also took advantage of the fact that in turtle utricle, type I hair cells have a sharply restricted distribution (Jorgensen 1974Go, 1988Go; Moravec and Peterson 2004Go; Xue and Peterson 2006Go) and significantly more stereocilia than neighboring type II hair cells (Moravec and Peterson 2004Go) to estimate differences between the arrays of different hair cell types. We presented some of these data in abstract form (Peterson and Rowe 2001Go).


    METHODS
 TOP
 ABSTRACT
 INTRODUCTION
 METHODS
 RESULTS
 DISCUSSION
 GRANTS
 ACKNOWLEDGMENTS
 REFERENCES
 
Seven juvenile turtles, Trachemys (Pseudemys) scripta elegans, of both sexes (3.5- to 5-in carapace length; Kons Scientific, Germantown, WI) provided useful data. We used two turtles for a low-magnification view of the macula (Fig. 2) and images of intact bundles (GoGoGoGoGoGoGoGoGoFig. 12) and five turtles for quantification (Table 1): autocorrelation analysis of a medial-to-lateral transect (U5, U50) and counts of total utricular bundles (U5 and 3 additional turtles). Animal care protocols are published (Brichta and Peterson 1994Go). We killed all turtles with Euthasol (390 mg pentobarbital sodium and 50 mg phenytoin sodium/ml; 0.5 ml im) and followed Ohio University Animal Care and Use Committee guidelines in all experiments.


Figure 2
View larger version (176K):
[in this window]
[in a new window]
 
FIG. 2. Scanning electron micrograph of left turtle utricle. It has been sonicated to remove the otoconial membrane and hair bundles. Striolar bundles, which have relatively large apical surfaces, form a crescent-shaped band that parallels the lateral margin of the macula ({blacktriangleup}). The macula can be subdivided into 4 zones based on differences in bundle structure, afferent terminal morphology, and calretinin-immunoreactivity. Zones 1 and 4 correspond to lateral (LES) and medial (MES) extrastriolae, respectively. Zones 2 and 3 are parallel bands that collectively form the striola. {leftrightarrow}, orientation of the radial transects analyzed in this study.

 

Figure 3
View larger version (75K):
[in this window]
[in a new window]
 
FIG. 3. Scanning micrograph showing striolar arrays from 1 transect. Some arrays from adjacent lateral and medial extrastriolae are also visible at top and bottom, respectively. - - -, line of polarity reversal. At present, we define zone 1 (LES) as hair cells lateral to the line of polarity reversal. Their arrays tend to be small and round. Zone 2 is a band of type II hair cells; it is ~15 µm wide in scanning micrographs. Zone 3 is a band of type I hair cells and any interspersed type II hair cells (Moravec and Peterson 2004Go; Xue and Peterson 2006Go). Collectively, zones 2 and 3 form the striola. Zone 4 (MES) is defined as hair cells medial to zone 3. Zone 4 arrays become increasingly long and narrow toward the medial macula. Yellow highlighting: bundles with more than 77 stereocilia, i.e., probable type I hair cells. Two bundles with 80–85 stereocilia are closer to the reversal line than expected. These may or may not be incorrectly categorized as type I hair cells. In confocal material, we occasionally see identified type I hair cells very close to the reversal line (see Fig. 1C in Moravec and Peterson 2004Go). Some probable type I hair cells (based on stereocilia counts) were not considered measurable (examples marked with asterisks). Note that for probable type I hair cells, stereocilia number increases and center-to-center spacing decreases with distance from the reversal line.

 

Figure 4
View larger version (35K):
[in this window]
[in a new window]
 
FIG. 4. Zonal variation in bundle dimensions with distance from the line of polarity reversal (0 on abscissa). A, C, and E: in these scatter plots, each symbol represents 1 measured array. Black symbols: U5; gray symbols: U50. Insets: measured variables. Legend in A applies to C and E. Array area is markedly greater in the striola than in the extrastriolae (A) because striolar arrays are very wide (C). Striolar array length (E) is similar to that in the MES. Arrays in U50 are generally larger than those in U5, especially in the extrastriolae. They show the same spatial patterns except that array length in U5 (but not U50) decreases just medial to the striola before rising to MES levels. Loess fits summarize spatial trends in the variables for U5 (dark blue lines) and U50 (light blue lines). Small rectangles indicate approximate boundaries of striolar zones 2 and 3. Large symbols represent arrays with more than 77 stereocilia, i.e., probable type I hair cells. Two MES arrays have 78 stereocilia; they are misclassified by the criteria used here because there are no type I hair cells in the MES. B, D, and F: box plots summarize differences between zones and hair cell types for U5 and U50 combined. Array areas and widths in zone 2 are larger than those of other presumptive type II hair cells and not significantly different from those of type I hair cells. In contrast, array lengths are comparable for all hair cells medial to the reversal line. Arrays were assigned to hair cell type and zone based on location and stereocilia number (see Zonal variation). Circle: median. Gray boxes: interquartile range. Black, notched boxes: confidence intervals of the median. Nonoverlapping confidence regions indicate significant differences. We confirmed this using robust analogs of 1-way ANOVA and post hoc comparisons. Whiskers: 1.5 x interquartile range. Isolated points: outliers.

 

Figure 5
View larger version (41K):
[in this window]
[in a new window]
 
FIG. 5. Variation in stereocilia number (AC) and spacing (DF) with distance from the LPR. Stereocilia number and spacing are approximately constant in most of the MES and LES (A and D), but they exhibit marked spatial gradients within the striola (B and E). Stereocilia number increases from lateral to medial for most hair cells (B), and stereocilia spacing decreases (D). Regression lines in B and D summarize spatial trends in the variables for U5 (dark blue lines) and U50 (light blue lines). Equations are robust MM regression equations (P < 0.001). Other plotting conventions as in Fig. 4. In B and E, 4 arrays with >77 stereocilia appear to be in zone 2 (i.e., within 15 µm of the reversal line). They may or may not be misclassified; type I hair cells occasionally occur very close to the LPR. C and F: bubble plots showing zonal differences in stereocilia number and spacing for U5 hair cells in the striola and adjacent extrastriolae. Coordinates represent distance from the upper left corner of the transect. Each symbol represents 1 measured array. Symbol size is scaled to the number (C) and center-to-center spacing (F) of stereocilia. Note the reciprocal relation between stereocilia number and spacing in striolar zones 2 and 3.

 

Figure 6
View larger version (16K):
[in this window]
[in a new window]
 
FIG. 6. Zonal variation in stereocilia number (A) and spacing (B) for U5 and U50 combined. The box plots summarize differences between predicted zones and hair cell types. Stereocilia numbers for probable type II hair cells are greater in the striola, especially zone 2, than in the extrastriolae, consistent with earlier results on identified type II hair cells (Moravec and Peterson 2004Go). Probable type I hair cells are defined as having more than 77 stereocilia. Stereocilia spacing is significantly greater in striolar zone 2 than elsewhere in the macula (B).

 

Figure 7
View larger version (39K):
[in this window]
[in a new window]
 
FIG. 7. Variation in hair cell orientation along a radial transect through the macula. The vector plots contrast orientation of hair cell apical surfaces (left) and hexagon axis 1 (probable activation axis of the bundle; right). Each arrowhead represents 1 measured bundle in U5. Orientation of the arrowhead reflects alignment of the apical surface or hexagon axis. Coordinates represent distance from the upper left corner of the transect. The LPR is near –70 µm on the ordinate. Apical surfaces appear to be well aligned, but the orientation of hexagon axis 1 is markedly irregular.

 

Figure 8
View larger version (17K):
[in this window]
[in a new window]
 
FIG. 8. Histograms compare the distributions of 4 indicators of hair cell orientation. The ordinate indicates percent of total cells at each orientation. Distributions for orientation of the apical surface (A) and axis of bilateral symmetry (B) are narrow: ~2/3 of total bundles are oriented within ±10° of the transect line. The orientation of hexagon axis 1 is significantly broader, whether one includes all bundles (C) or eliminates bundles whose activation axes are most likely to be misidentified (i.e., bundles in which axis 1 is >20° from the ABS; D).

 

Figure 9
View larger version (23K):
[in this window]
[in a new window]
 
FIG. 9. Dot plots show the orientation of hexagon axis 1 (relative to mathematical 0) for samples of radius 20 µm taken at each of 8 (U5) or 9 (U50) locations along the transect. Each symbol represents one hair bundle. Dashed vertical line: mean orientation for all bundles. Vertical dotted lines: orientations ±25° from the mean orientation. Most bundles fall within this 50° spread, but the distributions at each location are broad. Numbers in parentheses at right show the number of measured bundles at each location.

 

Figure 10
View larger version (18K):
[in this window]
[in a new window]
 
FIG. 10. Simulation of afferent directional tuning profiles (red) for a sample of 16 hair cells (gray curves) found within a circle of 20-µm radius at position 3 on the U5 transect. Simulations are run for 3 values of I/E asymmetry: 0.1 (A), 0.2 (B) and 0.5 (C). Signals from individual hair cells are represented by asymmetrical cosine functions normalized to the peak response; afferent profiles are also normalized. Dark horizontal dotted line: 0 response magnitude. 0 on the abscissa: peak excitatory stimulus orientation. Vertical gray dotted lines: stimulus orientation orthogonal to the peak excitatory axis. Horizontal gray dotted line: amplitude of the afferent responses for directions orthogonal to the peak (± 90°), as a percentage of the peak response.

 

Figure 11
View larger version (20K):
[in this window]
[in a new window]
 
FIG. 11. Effect of sample size on the range of hair cell activation axes and simulated afferent directional tuning in U5 (gray) and U50 (black). Samples estimate the collecting areas of afferent terminals. Hair cell activation axes are estimated from the orientation of hexagon axis 1. A: range of hair cell activation axis orientations increases with sample radius, approximately doubling as sample radius increases from 10 to 50 µm. Symbols: average values. Bars: 1 SD from the mean (drawn in one direction only, for clarity). B: afferent orthogonal response magnitude, as a percentage of maximum response, also increases with larger radii but much more gradually. For both utricles, orthogonal response magnitude decreases as asymmetry in hair cell responses increases (~12% from 0.1 to 0.2; ~38% from 0.2 to 0.5); this increase is independent of collecting area radius. Plotting conventions as in A. C: dependence of bundle orientation range and afferent orthogonal response on sample radius. Data from A and B are expressed as a percentage of the maximum response for direct comparison. The range of bundle activation axis orientations approximately doubles between radius values of 10 and 50 µm, while orthogonal response magnitudes increases by 20–25%.

 

Figure 12
View larger version (64K):
[in this window]
[in a new window]
 
FIG. 12. Differences in the structure of hair bundles in the medial extrastriola (A) and striola (B and C). The upper image in each panel shows representative hair bundles; below the micrographs are simplified drawings of each bundle type in lateral view along with the associated arrays. In the lateral views, only a few rows of stereocilia are depicted for clarity; heights of kinocilium and stereocilia are from Xue and Peterson (2006)Go. The arrays are tracings from micrographs of sonicated bundles. A: bundles in the medial extrastriola (MES) have long narrow arrays. Stereocilia are short (relative to the kinocilium) and the height step between adjacent stereocilia is very small. B: type I bundles (striolar zone 3) have wide arrays with numerous, closely packed stereocilia. The kinocilium is shorter than on MES bundles and stereocilia are taller, with larger steps between adjacent stereocilia. In the micrograph, the type I hair cell (I) can be identified because the epithelium was fractured to reveal its globular soma. The calyx is visible as debris on the soma surface (Lapeyre et al. 1992Go). A cylindrical type II hair cell (II) appears to wrap around the type I soma. Confocal micrographs of identified hair cells in utricular slices indicate that this is a very common configuration in striolar zone 3 (Xue and Peterson 2006Go). C: bundles in striolar zone 2 are homogeneous, with wide arrays and few, widely spaced stereocilia. The tallest stereocilia equal the kinocilium in height, and the height step between adjacent stereocilia is very steep. Center-to-center spacing of stereocilia is wider than elsewhere in the macula (Fig. 6B); this probably reflects the fact that stereocilia shaft diameters are also significantly greater in this zone (E. H. Peterson and W. J. Moravec, unpublished). D: schematic diagram (redrawn from Nam et al. 2005Go) illustrating some mechanical consequences of different bundle configurations. Simulations suggest that shear flow of endolymph (arrows) produces fluid drag on stereocilia, thereby tensing gating springs and opening transduction channels (Nam et al. 2005Go). This fluid drag is proportional to the height step between adjacent stereocilia ({Delta}ySS; because large height steps increase the area for fluid drag that tensions gating springs) and the height at which fluid force is applied to each stereocilium (yi; because shearing force increases with distance from the apical surface). The black arrow shows one example. Both {Delta}ySS and yi increase with proximity to the LPR (from A to C), suggesting that fluid drag will become increasingly effective. Because time to peak channel activation is much faster in fluid-forced bundles than when force is applied to the kinocilium via its attachment to the otoconial membrane (Nam et al. 2005Go), we anticipate that striolar bundles (B and C) will respond more rapidly than MES bundles, enabling them to signal higher frequency head transients and leading to a relative phase lead when stimulated by sinusoidal stimuli.

 

View this table:
[in this window]
[in a new window]
 
TABLE 1. Macular area, total hair cells, and number of measured hair cells

 
Scanning electron microscopy

We perfused killed turtles transcardially with oxygenated turtle Ringer solution (Hounsgaard and Nicholson 1990Go) followed by fixative: 2% glutaraldehyde, 4% paraformaldehyde, 2 mM MgCl2 in 0.1 M sodium phosphate buffer, pH 6.2 to visualize intact bundles, or 5% glutaraldehyde, 4% paraformaldehyde, 2% sucrose, 0.2 M picric acid, 0.125 M phosphate buffer, pH 7.4 to visualize arrays. We dissected utricles in fresh fixative, postfixed them in 2% OsO4 for 1 h, followed by rinsing (10 times in distilled water), osmium-thiocarbohydrazide (OTOTO) treatment (Furness and Hackney 1986Go), dehydration (in graded ethanol series), critical point drying, sputter coating with gold palladium, and examination using a Jeol JSM-840 or a Zeiss DSM 962. Prior to post fixation we treated utricles in one of two ways. To visualize intact bundles, we removed the otoconial membranes using fine forceps. For autocorrelation analyses of stereociliary arrays, we removed otoconial membranes and hair bundles by sonication (30–60 s in 70% ethanol) or with a fine (0000) sable brush under high magnification, and we oriented the macula so that the viewing angle was normal to the transect. Arrays were considered measurable if they were intact and the hair cell surface was free of fissures, debris, and stereocilia.

Autocorrelation analysis: quantification of arrays

We quantified the number and arrangement of stereocilia as follows. We photographed arrays at x5,000 and quantified features of interest using autocorrelation analysis. The term array refers to the number, spacing, and distribution of stereociliary remnants on the hair cell's apical surface (Fig. 1B). We published details of this method previously (Rowe and Peterson 2004Go). Briefly, we used a custom program running under MATLAB (ver. 7; MathWorks) to record data from each hair cell (location of stereocilia and kinocilium, perimeter of apical surface and array) and to derive the following hair cell and bundle descriptors (see sections 2.2 and 2.3 and Figs. 1 and 2 in Rowe and Peterson 2004Go).

  1. ) Area, orientation, and shape of the apical surface. For each hair cell, we fit an ellipse to the apical surface perimeter and used parameters of the fitted ellipse to specify the area, long and short axes, and orientation (relative to the axis of the transect) of the apical surface.
  2. ) Area, long and short axes, and orientation of the stereociliary array were quantified using the same methods as for the apical surface.
  3. ) Stereocilia number.
  4. ) Stereocilia spacing. Average center-to-center spacing for each bundle is given by the initial peak in its autocorrelogram (see Fig. 2 in Rowe and Peterson 2004Go).
  5. ) Axis of bilateral symmetry (ABS). Mean x and y coordinates of all stereocilia in an array estimate the array center. The orientation of a vector that originates at the array center and terminates at the kinocilium defines the ABS.
  6. ) Estimated activation axis relative to ABS. Arrays are approximately hexagonal; thus they present three axes. Tip links generally run along the axis closest to the ABS (axis 1), so this is the best estimate of the bundle's activation axis (Pickles and Corey 1992Go). It also specifies whether bundle geometry is "loose" or "tight" (Bagger-Sjöbäck and Takumida 1988Go; see Fig. 8 in Rowe and Peterson 2004Go).
  7. ) Spacing slope is the change in average stereocilium spacing with distance from the kinocilium, measured parallel to the ABS. Slopes are negative if stereocilia spacing decreases toward the short end of the bundle.

Zonal variation

Our previous work indicates that the utricular macula in T. scripta can be divided into four zones (Figs. 2 and 3) based on differences in bundle structure, afferent terminal morphology, and calretinin-immunoreactivity of hair cells and afferents (Moravec and Peterson 2004Go; Xue and Peterson 2006Go; Xue et al. 2005Go). Zones 1 and 4 correspond to the lateral (LES) and medial (MES) extrastriolae, respectively. Zones 2 and 3 form the striola. In undehydrated utricles, zone 2 is ~20 µm wide. Zone 3 is a 50- to 60-µm-wide band of type I hair cells and any interspersed type II hair cells (Moravec and Peterson 2004Go; Xue and Peterson 2006Go).

To assess zonal variation in array structure, we compared arrays along a medial-to-lateral transect that spans all four zones (Fig. 2, {leftrightarrow}). Hair bundles along this transect have similar activation axes. Thus we were able to analyze zonal variation while holding directional variation nearly constant. Our previous work suggests that other aspects of bundle structure do not differ significantly with transect orientation (stereocilia number: Moravec and Peterson 2004Go; bundle heights: Xue and Peterson 2006Go); therefore zonal variation along the transect used in this study is likely to be representative of zonal variation along any transect.

We illustrate zonal variation in three ways. Scatter plots (Figs. 4, A, C, and E, and 5, A, B, D, and E) show how variable values change with distance from the line of polarity reversal (LPR). Zero on the abscissa represents a straight line fitted to the irregular trajectory of the LPR. Hair cell position is the perpendicular distance from this fitted reversal line to the hair cell's kinocilium. Both turtles showed similar spatial trends, but we plot data for U5 and U50 separately because their values on some variables at some locations (usually the extrastriolae) are significantly different. Box plots (Figs. 4, B, D, and F, and 6) summarize broad differences between zones and hair cell types for the two utricles combined. Finally, bubble plots (Fig. 5, C and F) give semi-quantitative overviews of reciprocal patterns of stereocilia number and spacing within the striola.

To categorize data for box plots, we approximated hair cell type and zonal boundaries as follows. Type I hair cells can generally be identified because of their high stereocilia counts. In a previous study (Moravec and Peterson 2004Go), the highest number of stereocilia on identified type II hair cells was 77. Therefore we designated all cells with more than 77 stereocilia as probable type I hair cells. There are two possible sources of classification error. A small number of identified type I hair cells (5/94) (Moravec and Peterson 2004Go) had fewer than 78 stereocilia, so some zone 3 bundles labeled type II in the present study may have been type I hair cells. In addition, apparently immature bundles (Severinsen et al. 2003Go) in zone 3 were assigned to hair cell type based on stereocilia counts, so some may have been immature type I hair cells. Such cells were very rare.

Type I hair cells and any intercalated type II hair cells form zone 3 (Moravec and Peterson 2004Go; Xue and Peterson 2006Go). This zone begins ~15 µm medial to the LPR and is 40–50 µm wide, which suggests a shrinkage factor (compared with undehydrated confocal material) of ~1.25x. Two cells with >77 stereocilia were within 15 µm of the reversal line. This is consistent with confocal material showing occasional, identified type I hair cells very close to the LPR (Moravec and Peterson 2004Go). The other zonal boundaries were as follows. Hair cells lateral to the reversal line form zone 1 (LES); those between the reversal line and zone 3 form zone 2; those medial to zone 3 form zone 4 (MES). Resulting sample sizes for each zone and hair cell type are given in Table 2.


View this table:
[in this window]
[in a new window]
 
TABLE 2. Samples for height measurements: number of measured bundles

 
Statistical analysis of zonal variation

To facilitate comparisons with published studies using scanning electron microscopy (SEM), we made no correction for shrinkage. Exploratory and inferential statistics were implemented in Statistica (ver. 7.1; StatSoft) or S+ (ver. 7; Insightful). We used robust statistics for our analyses (Wilcox 2005Go) because many of our variable distributions were nonnormal (see DISCUSSION in Xue and Peterson 2006Go). Briefly, they are powerful but make few assumptions about the shape of underlying variable distributions. We implemented robust functions in S+. To compare bundles in different zones we used t1way (analog of 1-way ANOVA) and lincon (for multiple comparisons; Wilcox 2005Go; Chapter 7). To describe spatial gradients in and near the striola (Fig. 5, B and E), we used robust MM linear regression (available in S+). We summarized spatial trends across the macula using Loess fits (a form of smoothing using weighted, local regressions on data that cannot be fit with simple linear or quadratic equations) (Cleveland 1993Go). For the fits in Figs. 4, A, C, and E, and 5, A and D, we used a local quadratic fit (because there are local maxima and minima in our data), a Gaussian weighting function, and a span (which determines the degree of smoothing) of 0.1–0.2. We implemented Loess fits in S+.

Directional specificity

We also quantified the random variation in bundle orientation. We used one hexagon axis (axis 1, i.e., the hexagon axis closest to the ABS) to estimate each bundle's activation axis, and we examined variability in axis 1 orientation for all bundles along the transect (see Directional specificity, Radial populations) and for bundles in small, circular areas with dimensions similar to the collecting areas of single utricular afferents (Directional specificity, Local populations). We used the small samples to examine two questions. What is the variation in axis 1 orientation for a small group of hair cells, such as might provide input to a single afferent (Fig. 1A, a)? How does variability in axis 1 orientation increase as sample area increases (Fig. 1A, b)? Answers to these questions may help explain some directional tuning characteristics of utricular afferents (Dickman et al. 1991Go; Fernandez and Goldberg 1976Go; Si et al. 1997Go).

To estimate variation in hair cell activation axes that exists within the collecting area of a typical afferent, we sampled 8 (U5) or 9 (U50) locations evenly spaced along the medial-to-lateral transect. Figure 1A shows approximate locations of the eight samples for U5. At each location, we simulated the collecting area of an afferent as a circle with a radius of 20 µm. This approximates the average size of utricular afferent terminals (Table 3). At each location, we also created ≤24 additional samples arranged in two concentric rings around the central sample (Fig. 1A, a). The inner ring comprised nine circular samples spaced at 45° angles around the central sample and displaced from it by one radius (20 µm). The outer ring comprised 15 circular samples, spaced at 22.5° angles around the central sample and displaced from it by two radii (40 µm). We eliminated samples that impinged on the borders of the transect, resulting in a total of 319 samples for the two utricles. These samples provided a robust estimate of the variation in bundle orientation to be expected within the collecting area of single utricular afferents.


View this table:
[in this window]
[in a new window]
 
TABLE 3. Collecting areas of utricular afferents

 
For each of the samples (≤25) at each position along the transect, we recorded axis 1 orientation for all measured hair cells within the 20-µm-radius circle. We used these data to examine effects of variation in hair cell activation axes on afferent directional tuning. To do this, we simulated individual hair cell responses as the sum of two cosine functions, R({theta}) = cos({theta}) + aH[cos({theta} + {pi})], where {theta} is the angle between stimulus direction and the hair cell excitatory axis, H[cos({theta} + {pi})] is a half-wave rectified cosine phase shifted 180°, which partially cancels the negative half cycle of the first term (required because hair cell responses to excitatory and inhibitory stimuli are not symmetrical), and a is the asymmetry factor. Values for a were chosen to yield inhibitory/excitatory (I/E) ratios of 0.1, 0.2, or 0.5. The lower ratios correspond to values for hair cells reported in the literature (Holt et al. 1997Go; Hudspeth 1983Go; Shotwell et al. 1981Go; Vollrath and Eatock 2003Go). The upper bound of 0.5 takes into account the possibility that values in the literature are underestimates (Eatock 2000Go; Ricci et al. 1998Go). For each of the 319 samples, we simulated an afferent directional tuning curve by summing all hair cell profiles in that sample. Summation was linear and all hair cells were equally weighted. This is a reasonable approximation because local populations of hair cells in turtle utricle tend to be homogeneous. The only exception is a narrow band (40–50 µm wide), zone 3, where type I and II hair cells occur side by side. To the extent that single afferents summate signals from hair cells that vary in bundle structure or in gain or response dynamics (as they do in turtle posterior canal) (Brichta et al. 2002Go), the situation might be different, but this remains to be determined empirically. Optimal orientation of the afferent was determined from the peak of the resulting tuning curve. Responses orthogonal to the optimal orientation were determined from the height of the tuning curve at points ±90° from the optimal orientation.

To explore the influence of afferent collecting area size on directional specificity, we repeated the analysis of (up to) 25 samples at each of eight to nine locations along the transect, except that we varied sample radius in 10 µm increments from 10 to 40 or 50 µm, depending on transect width (Fig. 1A, b). For radii >30 µm, we only analyzed the central sample at each location because more eccentric samples impinged on the borders of the transect.

Statistical analysis of directional specificity

To compare overall differences in the distribution of apical surface, ABS, and axis 1 orientations we used a Kolmogorov-Smirnov test. We used t1way and lincon to investigate the effect of increasing sample radius on variability of axis 1 angles (putative activation axes) and on orthogonal afferent response magnitude, assuming I/E asymmetries of 0.1, 0.2, and 0.5. To assess the effect of I/E asymmetry on simulated orthogonal response magnitude while holding sample radius constant, we used ancova (robust analysis of covariance, radius as covariate) (Wilcox 2005Go; section 11.8). Axis 1 ranges and orthogonal response magnitudes for U5 and U50 were not significantly different, so we collapsed them for analysis.

Hair bundle counts

We counted total bundles from one SEM montage and three utricular whole mounts stained with the f-actin probe phalloidin to visualize hair bundles (Table 1). Two utricles scanned at x10 magnification provided total bundle counts. A third, scanned at x40, provided total bundle counts and enabled us to distinguish hair cells medial and lateral to the LPR.


    RESULTS
 TOP
 ABSTRACT
 INTRODUCTION
 METHODS
 RESULTS
 DISCUSSION
 GRANTS
 ACKNOWLEDGMENTS
 REFERENCES
 
Figure 2 is a low-magnification scanning micrograph of the utricle of T. scripta that was sonicated to remove otoconial membrane and hair bundles. The fan-shaped macula forms a shallow bowl except posteriorly, where the neuroepithelium is more steeply curved. The striola is visible as a crescent-shaped band that parallels the lateral margin of the macula (arrowheads). There are ~8,500 total hair cells (Table 1); roughly 30% of these are lateral to the line of polarity reversal. Macular areas (Table 1) were measured using undehydrated whole mounts only (n = 2); they are slightly underestimated due to foreshortening of the curved macula.

We analyzed arrays from medial-to-lateral transects in two animals (U5, U50); the transects were 107 and 79 µm wide, respectively. Median hair cell number for the two transects was 765.5; of these, ~66% bore measurable arrays (Table 1). Table 2 summarizes the resulting data base. In this and other tables, we assigned arrays to macular zone and hair cell type as described in METHODS. Figure 3 shows examples of arrays analyzed in striolar zones 2 and 3 and adjacent LES (zone 1) and MES (zone 4). Putative type I hair cells (those having more than 77 stereocilia) are highlighted.

Zonal variation in stereociliary arrays

Array structure varies with medial-to-lateral position across the macula. Both utricles showed similar patterns, but length and area measurements from the extrastriola of U50 were sometimes larger than those in U5. This is probably due to differential shrinkage rather than developmental stage because U5 was slightly larger than U50 (Table 2). Array features that exhibit the most striking regional variation are illustrated in Figs. 46. Two other array features are not illustrated because they show little zonal variation: orientation of the presumptive activation axis (Axis 1 orientation) and changes in stereocilia spacing with distance from the kinocilium (Spacing slope). Table 4 shows summary statistics for all measured variables. Detailed tables are available as supplemental material.1


View this table:
[in this window]
[in a new window]
 
TABLE 4. Average differences in hair bundle structure

 
Array dimensions

Array area is highest in a band that begins at the LPR and extends 50–60 µm medially (Fig. 4A). Based on its location and stereocilia counts, this band corresponds to striolar zones 2 and 3 as identified in confocal micrographs (Moravec and Peterson 2004Go; Xue and Peterson 2006Go). For probable type II hair cells, array area is greatest in zone 2 (Fig. 4B, left). The high array areas in zone 3 (Fig. 4A) are due to bundles with very high stereocilia counts, i.e., probable type I hair cells (Fig. 4B, right). Zone 3 bundles with smaller array areas have lower stereocilia numbers; they are probably type II hair cells. Apical surface areas (not illustrated) are highly correlated with array areas (U5: r = 0.80, P < 10–6; U50: r = 0.63, P < 10–6).

Array width closely parallels array area (Fig. 4C), i.e., arrays are significantly wider in striolar zones 2 and 3 than in the extrastriolae. Correlations between array area and array width are 0.90 (U5, P < 10–6) and 0.66 (U50, P < 10–6). For type II hair cells, array width is largest in striolar zone 2 and smallest in extrastriolar zones 1 and 4 (Fig. 4D). Array widths of type II hair cells in zone 2 are not significantly different from those of probable type I hair cells (Fig. 4D).

Array length is less strongly correlated with array area (U5: r = 0.47, P < 10–6; U50: r = 0.32, P < 10–6). Instead of peaking in the striola as do array area and width, striolar array lengths are roughly equal to (U5) or slightly less than (U50) those in the MES (Fig. 4, E and F). Thus arrays in the striola are large because they are distinctively wide; they are not distinctively long. We consider the significance of wide striolar arrays in the DISCUSSION. Bundles in the LES (zone 1) have significantly shorter arrays than other macular bundles (Fig. 4F).

Stereocilia number and spacing

Stereocilia numbers, like array area and width, are highest in a band that begins at the LPR and is 50–60 µm wide, i.e., in the region we operationally define as the striola (Fig. 5A). Stereocilia numbers in the striola are significantly greater than those in the extrastriolae (zones 1 and 4), whether or not one includes putative type I hair cells in the striola-extrastriola comparison (Fig. 6A). This is consistent with results on identified type I and type II hair cells (Moravec and Peterson 2004Go). Stereocilia spacing is approximately uniform across the macula except for a narrow band just medial to the LPR (zone 2; Fig. 5D), where spacing is significantly greater than elsewhere in the macula (Fig. 6B).

Within the striola, stereocilia number and spacing change systematically from lateral to medial (Fig. 5, B and E). Stereocilia numbers increase (Fig. 5B) and spacing decreases (Fig. 5E). The dependence of stereocilia number on spacing is significant (MM regression: Wald test P = 6.7 x 10–7). Thus hair bundles in zone 2 have wider center-to-center spacing than other utricular bundles, and bundles in zone 3 have the highest stereocilia counts (Fig. 5, C and F). In addition, the number and spacing of stereocilia on presumptive type I hair cells (zone 3 only) show a significant dependence on position from the reversal line (MM regression: Wald test P = 1.8 x 10–5 and P = 1.6 x 10–9, respectively). Thus there are systematic spatial gradients in stereocilia number and spacing from lateral to medial margins of the striola.

Directional specificity

We asked two questions about the directional specificity of hair cells in the utricle. First, how precise is the directional specificity of a radial strip of hair cells that appear, grossly, to be aligned parallel to each other? Such radial populations are presumed to be important functional units of the utricle because they are "tuned" to the same direction of force by the parallel orientation of their activation axes. Second, how tight is the directional specificity of a small, local hair cell population, which might provide the input to a single afferent?

Radial populations

To assess directional specificity (tightness of bundle alignment) in a radial strip of receptors we examined hair cells of the transect. We compared three measures of hair cell orientation. 1) Apical surface orientation. This is the measure used to construct classical maps of hair cell orientation in otoconial organs, but it has only a loose relation to the activation axis (Rowe and Peterson 2004Go). 2) Axis of bilateral symmetry (ABS). The ABS is more functionally relevant than apical surface orientation because it runs from the kinocilium (the presumed site of force application) through the centroid of the bundle; but it is not equivalent to the activation axis unless it runs along the axis that carries the gating springs (elastic elements that help tense mechanotransduction channels when the bundle deflects toward the kinocilium). 3) Axis 1, the hexagon axis closest to the ABS. This is the best estimate of the hair cell's activation axis because it is most likely to carry the bundle's gating springs.

Vector plots confirm the subjective impression gleaned from scanning micrographs that the apical surfaces of hair cells in the transect are aligned approximately in parallel (except for a small group of cells at lower right in this utricle; Fig. 7, left). Alignment of the ABS is slightly less regular (not shown), and alignment of axis 1 is markedly irregular (Fig. 7, right). Distributions for apical surface area, ABS, and axis 1 relativeto the transect line are shown in Fig. 8, AC. The distribution for axis 1 (Fig. 8C) is significantly broader than for the other two variables (Fig. 8, A and B; Kolmogorov-Smirnov test for goodness of fit, P < 0.02). A nearly identical pattern was observed in a second utricle. This broad distribution of axis 1 relative to the transect line suggests that hair cells in a radial strip such as the transect are not as tightly tuned to a single direction of head movement as their apical surfaces suggest.

Dispersion of axis 1 around the transect line estimates variability in bundle activation axes relative to a single direction of force (i.e., a force parallel to the transect line). Three factors sum to produce this dispersion: orientation of apical surfaces relative to the transect line, orientation of the ABS relative to the apical surface, and orientation of the hexagonal array relative to the ABS. The latter variable measures whether the bundle is "loose" or "tight" (Bagger- Sjöbäck and Takumida 1988Go; Rowe and Peterson 2004Go). Multiple regression analysis suggests it is most important in producing the spread of axis 1; the squared semi-partial correlation, which gives the proportion of axis 1 variance uniquely accounted for by the looseness or tightness of the array was 0.66 (U5) and 0.67 (U50). Thus the dispersion of axis 1 around the transect line (Fig. 8C) arises primarily because the hexagonal array of stereocilia is rotated relative to the ABS.

The preceding analysis assumes that axis 1 corresponds to the hair cell's activation axis (the hexagon axis that carries the gating springs). This assumption is least likely to be true for tight bundles; in a perfectly tight bundle, the ciliary array is rotated such that the ABS is equidistant (30°) from two hexagon axes, either of which may carry the gating springs. Thus the greater the rotation of axis 1 from the ABS (≤30°), the greater the possibility that axis 1 is not the activation axis. To assess the effect of mis-identifying the activation axis we removed the "tightest" bundles (defined arbitrarily as bundles in which the ciliary array was rotated more than 20° from the ABS). This restricted distribution (Fig. 8D) is not significantly different from the total distribution of axis 1 (Fig. 8C; Kolmogorov-Smirnov test, P > 0.2 for both utricles), and it displays significantly greater dispersion than the distribution of apical surfaces or of the ABS (Fig. 8, A and B; Kolmogorov-Smirnov test, P < 0.01 for both utricles). Thus using a more conservative criterion for identifying activation axes does not change the conclusion that hair cells along the transect are less directionally specific than their apical surfaces suggest.

Local hair cell populations

It would also be useful to know the directional properties of hair cells that provide input to single afferents because this helps us understand the origin of afferent directional tuning. Two factors will introduce variability in axis 1 orientation.

First, axis 1 (putative activation axis) exhibits local, apparently random differences in orientation (Fig. 1B). Figure 9 shows the orientation of axis 1 for all measured hair bundles in eight (U5) or nine (U50) samples. Samples are equally spaced along the transect (center-to-center spacing ~60 µm); each sample is 40 µm diameter (average collecting area of utricular afferents; Table 3). Thus each sample approximates a local population of hair cells that could provide the input to a single afferent. These local populations have a preferred direction but are broadly tuned. Average orientation of all 40 µm-diameter samples combined is indicated by vertical dashed lines. Over 90% (U5: 90.2%, U50: 94.7%) of all bundles are oriented within ±25° of this population average (vertical dotted lines). For all 176 samples from U5 the mean range of orientations was 47.6 ± 10.14° (mean ± SD); for the 143 samples from U50, the mean range was 45.8 ± 13.33°.

To examine the effects of this variation in activation axes on afferent directional tuning, we represented individual hair cell responses as asymmetrical cosine functions. Figure 10 shows one example. For an I/E ratio of 0.2, mean orthogonal response amplitude for all samples combined was 8 ± 2.2% (mean ± SD) of the maximum response in one transect and 7.9 ± 1.9% in the second. Reducing the I/E ratio to 0.1 increased the mean values to 9 and 8.9%, respectively. Increasing the I/E ratio to 0.5 reduced the mean values to 5 and 4.9%.

A second factor affecting the directional specificity of hair bundle populations is the systematic directional variation in bundle orientation across different radial transects (Fig. 1A, b). One consequence is that the larger the collecting area of an afferent, the more likely it is to sample hair cells with different bundle orientations. To learn how the spread of bundle orientations changes with increased sample area, we varied the radius of all samples at each of the eight (U5) or nine (U50) epithelial locations from 10 to 40 µm (U50) or 50 µm (U5). Figure 11A illustrates results for the two utricles; they were not significantly different. A robust ANOVA on the two utricles combined indicated that there is a significant effect of sample radius on the range of axis 1 angles (P < 0.000001). Each 10 µm increment in sample radius through 40 µm produced a significant increase in the range of axis 1 angles; results for samples of 40–50 µm did not differ.

To explore the influence of collecting area radius on directional tuning of afferents, we repeated the analysis shown in Fig. 10 but with additional samples of radii 10, 30, 40, and 50 µm (U5 only). Results for the two utricles did not differ significantly (Fig. 11B). The magnitude of the simulated orthogonal response depended on whether we assumed a hair cell I/E response asymmetry of 0.1, 0.2, or 0.5 (ancova, radius as covariate; P < 0.01 for all radii). It also showed a significant overall dependence on sample radius (P = 0.0000014) for I/E asymmetries of 0.1 or 0.2, but the only significant difference was between samples of radius 10 µm and the other, larger samples. Sample radius had no effect for I/E ratios of 0.5. Figure 11C shows normalized data from A and B to facilitate their comparison. The simulated magnitude of orthogonal responses (circles) showed a weaker dependence on sample radius than did axis 1 range (squares).


    DISCUSSION
 TOP
 ABSTRACT
 INTRODUCTION
 METHODS
 RESULTS
 DISCUSSION
 GRANTS
 ACKNOWLEDGMENTS
 REFERENCES
 
This study describes spatial patterns in the ciliary arrays of utricular hair bundles, i.e., the number, spacing, and distribution of stereocilia on the hair cell surface. These data are important for the light they may shed on functional differences between hair cells at different macular loci, and they provide essential input for computational models of utricular hair bundles. In addition, we took advantage of known differences in the location and stereocilia numbers of type I and type II hair cells in turtle utricle to estimate differences in array structure of different hair cell types. Our most important results are that 1) of the features measured, array width is the single most distinctive feature of striolar hair bundles, 2) within the striola there are negatively correlated gradients in stereocilia number and spacing, and 3) neighboring arrays vary 45–50° in orientation. We consider the functional implications of these results in the following text.

Methodological issues

Two factors affect the accuracy of our results. Stereocilia are generally easy to distinguish from microvilli (Peterson et al. 1996Go). But in a narrow band just medial to zone 3, the distinction was often difficult at the short end of the bundle. In this juxtastriolar region, heights of the tallest stereocilia increase toward striolar values (Xue and Peterson 2006Go). Thus stereocilia at the short end of the bundle may be difficult to identify because they are being resorbed as stereocilia at the tall end of the bundle elongate; such redistribution of actin occurs in developing chick auditory bundles (Tilney et al. 1992Go). One possibility is that we saw no dip in array length in the juxtastriolar region of U50, as we did in U5 (Fig. 4E) and utricular slices (Xue and Peterson 2006Go), because we incorrectly identified microvilli (as stereocilia) at the short end of the bundle. Alternatively U50 could represent a slightly different developmental stage. Our result that array lengths in the MES are as long or longer than those in the striola is consistent with conclusions from Xue and Peterson (2006)Go.

For analysis of directional specificity, we assumed that the hexagon axis closest to the ABS (axis 1) is the hair cell's activation axis. In intact bundles, the activation axis bears tip links (Gillespie et al. 2005Go; Nicolson 2005Go). Considerable evidence suggests that tip links typically (Pickles and Corey 1992Go; Pickles et al. 1991Go) follow a single axis of the bundle's hexagonal array, although anomalously oriented tip links have been described (Bagger-Sjöbäck and Takumida 1988Go; Hackney et al. 1988Go; Pickles et al. 1989Go). In auditory hair bundles (e.g., Pickles et al. 1989Go; Tilney et al. 1992Go) and frog saccule (e.g., Jacobs and Hudspeth 1990Go), this axis tends to parallel the ABS, i.e., the arrays are "loose." But in some "tight" vestibular bundles (Bagger-Sjöbäck and Takumida 1988Go; Flock 1964Go), arrays are rotated ≤30° relative to the ABS. The greater the rotation, the greater the uncertainty in identifying the activation axis. To gauge the effect of this uncertainty, we compared variability in estimated activation axes for all arrays versus arrays rotated only 0–20° from the ABS, i.e., we removed the "tightest" bundles. The two distributions were not significantly different (Fig. 8, C and D). Thus it is unlikely that errors in identifying activation axes of the tightest bundles, if any, would change our results.

Relation to previous work

Numerous studies have visualized ciliary arrays on inner ear and lateral line hair cells using light microscopy (e.g., Engström et al. 1962Go; Lindeman 1969Go), freeze fracture (e.g., Favre et al. 1986Go; Jacobs and Hudspeth 1990Go), transmission (e.g., Flock 1964Go; Flock and Wersäll 1962Go; Hackney et al. 1993Go; Morita et al. 1997Go), or scanning electron microscopy (e.g., Lim 1971Go; Platt and D'Andrea 1982Go; Severinsen et al. 2003Go; Tilney and Saunders 1983Go). Few studies quantified the images (Platt and D'Andrea 1982Go; studies reviewed in Jacobs and Hudspeth 1990Go; Morita et al. 1997Go; Tilney and Tilney 1988Go). To our knowledge, only one other study quantified zonal differences in utricular arrays (Platt and D'Andrea 1982Go).

Severinsen et al. (2003)Go described the development of turtle utricle. Total hair cell counts in our material are consistent with their predictions from macular area (see their Fig. 6C). They subdivided the macula into three zones (striola, MES, and LES) and equated the striola with the band of type I hair cells. Subsequent work indicates that the striola in turtle utricle is divided into two bands (zones 2 and 3), which differ in stereocilia number and spacing (Moravec and Peterson 2004Go; present results), bundle heights (Xue and Peterson 2006Go), and calretinin immunoreactivity of type II somata (Xue et al. 2005Go). Identified type I hair cells occupy only one of these bands (zone 3). A similar displacement of type I hair cells from the LPR is observed in the maculae of birds (Jorgensen 1989Go; Jorgensen and Andersen 1973Go; Rosenhall 1970Go; Si et al. 2003Go).2

Zonal variation

WHICH FEATURES DISTINGUISH STRIOLAR BUNDLES? Striolae are commonly described as a band of "big" bundles (e.g., Platt 1983Go; Rosenhall 1970Go). In what sense are they big? Probably not in heights. Kinocilia typically, and the tallest stereocilia frequently, are shorter in the striola than the extrastriola (e.g., Fontilla and Peterson 2000Go; Jorgensen 1988Go, 1989Go; Jorgensen and Christensen 1989Go; Lapeyre et al. 1992Go; Lewis and Li 1975Go; Lim 1977Go; Platt 1993Go; Rosenhall 1970Go; Xue and Peterson 2006Go). Apical surface areas of striolar bundles tend to be "large" (e.g., Lewis and Li 1975Go; Lindeman 1969Go; Lindeman et al. 1973Go; Severinsen et al. 2003Go). In turtle utricle, this is because they are wider than extrastriolar bundles (Jorgensen 1974Go; present results). Wide striolar bundles (i.e., wide parallel to the LPR) have also been reported in bats (Kirkegaard and Jorgensen 2001Go) and rodents (Lindeman et al. 1973Go). Indeed, Lindeman et al. (1973)Go commented that striolar bundles "bear a striking resemblance to the hair bundles of the inner hair cells in the cochlea." Thus bundle width is a major distinguishing feature of striolar hair bundles.

The functional significance of wide striolar bundles is uncertain. One hypothesis is that they allow numerous stereocilia to be arranged in a shallow staircase (i.e., with few rows from short to tall ends of the bundle). This allows a significant height step between successive rows (Fig. 12). Finite element models suggest that a large height step may 1) maximize the ability of endolymphatic shear flow to tense gating springs, thereby 2) producing very rapid peak transduction currents that 3) enable wide striolar bundles to signal high-frequency head transients (Fig. 12D) (Nam et al. 2005Go). In mammalian cochleae, high-frequency auditory stimuli are detected by the wide, shallow bundles of inner hair cells, which are probably stimulated exclusively by endolymph flow (Fridberger et al. 2006Go; Nowotny and Gummer 2006Go; reviewed in Robles and Ruggero 2001Go).

INTERNAL ORGANIZATION OF THE STRIOLA. As a group, striolar bundles differ from extrastriolar bundles, but they are not homogeneous. Structural subtypes of striolar bundles have been described in anamniotes, which have only type II hair cells (Lewis and Li 1975Go; Platt 1983Go), and in birds (Jorgensen 1989Go) and mammals (Lim 1977Go), where they were attributed to differences in hair cell type. In turtle utricular striola, the striking inverse relation between stereocilia number and spacing (Fig. 5, B and E) parallels gradients in bundle heights (Xue and Peterson 2006Go) (Figs. 6C and 7). These trans-striolar gradients clearly reflect differences in striolar zones. For example, ciliary spacing in zone 2 is greater than in zone 3; confocal images suggest this is because stereocilia shaft diameters are larger (E. H. Peterson and W. J. Moravec, unpublished). Trans-striolar gradients also reflect differences between type I hair cells within zone 3 (Fig. 5, B and E). Interestingly, type I cells contacted by calretinin-immunoreactive calyces (probable calyx afferents) (Desai et al. 2005Go) and calretinin-negative calyces (dimorphic afferents) occur laterally and medially, respectively, in zone 3, and these subgroups of type I hair cells differ significantly in bundle heights (Xue et al. 2005Go).