Abstract
Purpose:
To test the hypothesis that human, monkey, pig, sheep, cow, and goat eyes exhibit circumferential, radial, and interweaving collagen architecture in the posterior sclera.
Methods:
We analyzed 1,327 cryosections from the posterior poles of 4 human, 4 monkey, 5 pig, 8 sheep, 1 goat, and 2 cow eyes. Images were acquired using polarized light microscopy and processed to obtain polar fiber orientations relative to the center of the canal. Circumferential, radial, and interweaving regions were identified and analyzed for mean fiber orientation and anisotropy and region width and thickness.
Results:
Every eye exhibited circumferential, radial, and interweaving fibers in consistent locations. Radial fibers extended out from near the canal into the peripapillary and peripheral sclera in the innermost sclera. Circumferential fibers were directly adjacent to the canal and most prevalent in the outermost, posterior sclera. Interweaving fibers were found throughout the sclera thickness. Across all species, median anisotropy in the radial, circumferential, and interweaving regions were 0.95, 0.96, and 0.28, respectively.
Conclusions:
Regions of radial, circumferential, and interweaving fibers occur in the posterior pole sclera of human, monkey, pig, sheep, cow, and goat eyes. The consistency across species in scleral architecture suggests that they are primary organizational components whose functions should be better understood.
Collagen architecture heavily influences the properties of the sclera and, therefore, the biomechanics of the eye.
1,2 To understand eye physiology and pathophysiology, it is necessary to characterize the collagen architecture in the globe and, in particular, the posterior pole. Several techniques have been leveraged toward this, including wide-angle x-ray scattering (WAXS),
3–5 small angle light scattering (SALS),
6,7 magnetic resonance imaging (MRI),
8,9 and second harmonic generated imaging.
10 We have previously used polarized light microscopy (PLM) to visualize and quantify the collagen in the posterior sclera
11 and to identify the major organizational components of the posterior sclera of sheep eyes.
12 A key finding of our studies was that the collagen fibers of the posterior pole and peripapillary sclera of sheep eyes are organized into three distinct regions: a region of highly aligned circumferential fibers adjacent to the canal; a region of highly aligned radial fibers in the most anterior, or innermost to the globe, sclera; and a region of interweaving fibers without a primary alignment orientation.
Our goal was to test the hypothesis that circumferential, radial, and interweaving regions are consistent features of the collagen organization of the posterior sclera. Specifically, we studied if this is the case in eyes from human, monkey, pig, sheep, cow, and goat.
The PLM images were processed as previously described
12,13 to obtain Cartesian orientation and “energy” images (
Fig. 2). The orientation images display the collagen orientation in each pixel and were used for all of the analyses. The “energy” images reflect the signal strength in each pixel and were used to better visualize structure by adjusting the pixel brightness of the orientation images, as previously described.
12,13 Unless stated otherwise, all orientation images are shown masked this way. Using this processing, the canal region appears darker than the sclera, mainly due to lower collagen density (
Fig. 2). To better illustrate the architecture of the ONH region, we also applied fiber “tracing” methods, as previously described
12 (
Fig. 3).
For analysis, the orientation information was converted from the initially obtained Cartesian coordinates into polar coordinates. This was achieved by first manually marking the scleral canal center in each of the images. Then, the Cartesian coordinates were transformed to polar coordinates relative to the center of the canal. This transformation also simplified discerning the radial, circumferential, and interweaving regions because the polar orientations of 0° correspond to radial fibers and orientations of 90° correspond to circumferential fibers. Furthermore, they are more visually identifiable because radial areas appear green, blue, and cyan; circumferential areas are magenta, red, and orange; and interweaving regions have many different colors interweaving throughout.
We analyzed fiber orientations on each section following the previously reported methodology.
12 Briefly, we manually marked rectangular areas (200–500 μm per side) in radial, circumferential, or interweaving regions. The marked areas, 2749 in total, were processed to compute the circular mean and anisotropy in each. The circular mean represents the primary fiber orientation in the marked area. The anisotropy represents the spread of orientations, indicating the degree of alignment around the primary orientation. The anisotropy values were normalized so that perfectly aligned fibers have an anisotropy of 1 and fibers evenly dispersed in all directions have an anisotropy of 0. These values were used to create regional angle (circular mean) and anisotropy plots for each of the three region types in order to assess the patterns associated with the visually identified regions.
Linear mixed effects (LME) models, accounting for autocorrelations, or dependent similarities, between measurements from the same species, eyes, and sections, were used to assess if either the average angle or anisotropy were significantly related to the region types (radial, circumferential, and interweaving) identified when marking the images. This was done both for the individual species and between all species included in this study.
Note that a subset of the measurements in sheep eyes had been previously reported as part of an earlier, more limited study.
12 To facilitate comparability with the other species, the sheep data were reprocessed and reanalyzed, extending the analyses to include the depth-related measurements in this work.
An automated segmentation method was developed to study the radial and circumferential regions in more detail. The segmentation used the anisotropy and circular mean to isolate the radial and circumferential fibers from all others. The neighborhood and threshold used for the results presented in the manuscript were identified in a preliminary sensitivity study. To do this, each pixel was assigned an anisotropy and circular mean value based on the orientation values of its “neighborhood” of 51 × 51 pixels. The segmentation then assessed these values to determine if a pixel was radial, circumferential, or neither. The segmentation first separated the highly aligned pixels, including radials and circumferentials, from the less aligned pixels. The highly aligned pixels were identified using an anisotropy threshold. This threshold varied between 0.8 and 0.9, depending on the species. The segmentation then classified the highly aligned pixels based on the circular mean. Guided again by the fiber orientation quantification, angular thresholds were set for the radial and circumferential regions. Based on the range of −90° to 90° that the circular mean values fell in, values within 30° of the perfect radial at 0° (angles between −30° and 30°) were classified as radial. Orientations within 30° of the perfect circumferential at −90° or 90° (angles smaller than −60°, or larger than 60°) were classified as circumferential.
The purpose of this study was to visualize and quantify the collagen architecture in the posterior pole of several large mammals and to test the hypothesis that three consistent fiber arrangement types, circumferential, radial, and interweaving, appear in all of the species studied. Using PLM, we analyzed 1327 sections from 4 human, 4 monkey, 2 cow, 1 goat, 5 pig, and 8 sheep eyes and assessed the mean fiber orientation and anisotropy of each. To the best of our knowledge, this is the first systematic quantification of scleral collagen orientation and anisotropy in multiple species. Two main results arise from this work: (1) radial, circumferential, and interweaving fibers were found in the posterior pole of every eye from every species; and (2) the radial, circumferential, and interweaving regions had consistent patterns through depth across species. Both of these will be discussed in more detail below.
Radial, Circumferential, and Interweaving Fibers Were Found in the Posterior Pole of Every Eye from Every Species
Radial, Circumferential, and Interweaving Fibers Had Consistent Patterns Through Depth Across Species
Across the six species we studied, the radial fiber regions were present, on average in the 19.8% anterior of the sclera, varying between 10.5%, and as much as 25.7%. Circumferential fibers formed narrow regions between the canal and the radial fibers in the most anterior sclera, widening in deeper sclera. They were widest between 60% and 80% of the sclera depth. The interweaving fibers were found throughout the sclera depth.
The presence and consistent distribution of the radial, circumferential, and interweaving regions in every eye studied suggests that they are primary organizational components of the posterior pole of large mammal eyes. As such, these fibers likely play a role in determining the biomechanics of the posterior sclera and ONH. Future studies should identify the roles of the radial, circumferential, and interweaving regions and their influence on the overall biomechanics, especially given that the various species have different vision, eye physiology, and susceptibility to disease.
A similar consistent pattern of collagen organization has been shown for the cornea, with more interwoven fibers in the anterior stroma (exterior), than in the posterior (interior).
4,14,15 The overlapping of the radial and circumferential regions in the anterior part of the sclera suggests that, while the radial and circumferential regions are distinguishable from one another, they may also be structurally and functionally linked. Within these sections of overlap, there are often areas where the radial and circumferential regions are right next to each other (
Fig. 13). It seems plausible that some collagen bundles might transition between radial and circumferential regions.
16 Further studies should explore this possibility, as it could have major implications in how these two regions function biomechanically, especially in the meso- and macroscale transmission of forces.
There have been limited data published about the radial region, but our results support what was found in earlier studies. In our previous study of sheep, we characterized the radial region and found that it is anterior to the circumferential region, extends from the canal boundary to section edge, and comprises 60–180 μm of the depth.
12 The radial region has also been reported in humans. Pijanka and colleagues
17 described detecting fibers primarily oriented radially in the innermost (anterior) one-third of the peripapillary sclera. More specifically, the radial fibers were found in the anterior 150 μm of the posterior sclera. Although the function of the radial region is unknown, potential functions could include providing support for the vasculature in the choroid, providing support for the nerves coming into the ONH from the fovea, ensuring that they are not subject to large longitudinal deformations, and providing support for the lamina cribrosa, preventing large bending deformations of the ONH.
The circumferential region that we observed was also reported in several studies on the posterior scleral architecture. Most often referred to as a circumferential ring, these fibers are highly aligned and oriented parallel to the scleral canal.
18 Several techniques have been used to study this ring. Inverse modeling, as done by Grytz and colleagues,
19 predicted the existence of the circumferential region. Circumferential fibers have also been reported using WAXS,
17 SALS,
6 second harmonic imaging,
10 scanning electron microscopy,
20–22 and immunofluorescence,
23–25 and in our own study using PLM.
12 It is generally believed that the circumferential region provides structural support to the ONH.
26–28
The interweaving region has also been studied using inverse modeling and SALS. Through their use of inverse modeling, Grytz and colleagues
28 predicted that the collagen within the sclera is mostly randomly orientated. Using SALS, Girard and colleagues
6 did not find a preferential orientation for the fibers in wide regions of the sclera. At this point, it is important to consider the nomenclature. Several previous studies referred to this region as being “random” and “isotropic,” while herein we have instead opted to refer to them as “interweaving.” PLM's higher spatial resolution compared with SALS enabled us to both quantify the low degree of anisotropy in these regions and visualize the interweaving of the fiber bundles. Quantitatively, the only established characteristic of the interweaving region is decreased anisotropy. At this time, it is not understood how the interweaving region is organized, but given the apparent organization of the radial and circumferential regions, it seems unlikely that the region would be truly “random.” It seems instead that it simply remains beyond our understanding. As with the other regions, further research is needed to understand its structure and role in globe biomechanics.
This study was able to build on the success of the previous work analyzing radial, circumferential, and interweaving fibers through use of PLM. PLM has high spatial and angular sensitivity and resolution over a wide field of view and over a wide range of connective tissue densities, making pixel-by-pixel calculation of fiber orientation possible throughout the sclera.
11 PLM also works well in thin sections.
11 This is in contrast with WAXS
18,29 and SALS
6 that can analyze thick sclera samples but may suffer from artifacts when measuring thin sections.
30 The ability of PLM to quantify thin sections allowed for a higher depth resolution when analyzing stacks compared with studies published using scattering techniques. In addition, PLM was well suited for our goals in this work because it is robust to imaging systems and magnifications.
11 This allowed us to scale the magnification to best suit the sections from a given eye and to use images collected on several systems, without adversely affecting the analysis. PLM, like all techniques based on sectioning, is subject to section deformations, which may be smaller when the sections are thicker. Another advantage of PLM is that it did not require tissue labels,
31,32 dehydrating,
33 or flattening
5,34 the samples, which may introduce artifacts.
There are also several limitations of this study. First, our analysis of fiber orientations is limited to two dimensions due to the histologic sectioning and the imaging technique. This means that the collagen architecture reported in this work is a simplification of what is actually an even more complex three-dimensional (3D) structure. This simplification is not unique to PLM and is actually a common concern with fiber-analysis methods, such as WAXS and SALS. Although we have recently demonstrated an extension of PLM that allows measuring the full 3D orientation of the fibers based on PLM,
35 we have not yet applied the methods to this dataset. Our goal in this study was to determine the existence of regions of radial and circumferential fibers, which could be achieved with the more traditional PLM.
Another consideration when interpreting the results presented is that the sections had a thickness of 16–30 μm. This is thick enough that within each pixel, there are probably multiple overlapping fibers, perhaps with different orientations. The PLM implementation used in this work provides a measure of the dominant fiber orientation in each pixel. The images analyzed were obtained from tissues that were histologically processed. This processing may have introduced artifacts, such as tissue shrinkage or distortion, from the fixation and sectioning. However, we have shown that the distortions and shrinkage effects due to this fixation and sectioning are minimal.
36 The eyes used in this study were not all fixed at the same IOP. These changes in IOP could potentially affect fiber crimp and thus orientation and anisotropy.
37–39 This is also an issue for studies using WAXS and SALS.
32 Comparing eyes at different IOPs, we observed that the microscale uncrimping at elevated pressure reduced anisotropy without affecting the mean fiber orientation (
Fig. 12). Interestingly, the effects of IOP on anisotropy were substantial in the circumferential fibers and essentially null in the radial ones (
Supplementary Materials).
Calculation of the polar angles used to identify the radial and circumferential fibers was done based on manual identification of the center of the canal. This poses a number of potential problems. First, the markings were made based on visual recognition of the borders of the canal and the surrounding anatomy. Although someone experienced in identifying the structures within the ONH did the marking, it is subjective. Second, the center of the canal was marked based only on its two-dimensional section; it did not incorporate the fact that the eye is a 3D structure. Furthermore, in cases where the canal is not circular, the fibers may not be oriented radially with respect to the center of the canal but instead to the canal edge.
Our analysis was based on the assumption that the measurements are only from collagen. The presence of other birefringent components could, however, introduce artifacts. Elsewhere, we have shown that our PLM signal is in good agreement with other measurements of collagen, such as those from autofluorescence.
12 PLM also cannot distinguish elastin fibers, which can be visualized through nonlinear microscopy.
40
We recognize that the discretization of fiber populations into radial, circumferential, and interweaving is a simplification. It is essential to continue the work and refine the analysis to identify other aspects of the peripapillary sclera fiber organization that may have been lost in this analysis. For example, we have proposed that in the peripapillary sclera are groups of fibers that are tangential to the scleral canal, potentially appearing as circumferential near the canal and as radial farther away.
16
We have, to the best of our knowledge, presented the first systematic quantification of collagen fiber architecture and assessment of patterns of organization in the posterior pole of multiple large mammal species. We have confirmed that in human, monkey, cow, goat, sheep, and pig, the peripapillary sclera and posterior sclera have three organizational components: circumferential fibers, radial fibers, and interweaving fibers. These different fiber types appear in distinct regions with respect to the canal and anterior-posterior position. Future studies should identify the roles of the radial, circumferential, and interweaving regions and their influence on the overall biomechanics, especially given that the various species have different vision, eye physiology, and susceptibility to disease. This would benefit from determining the location of the lamina cribrosa relative to these regions. Future studies should also determine if the radial fibers are present in consistent locations in nasal-temporal and inferior-superior directions as well as through depth and if they are present in mice, rats, guinea pigs, and other small mammals often used in studies of the posterior pole.
Supported by National Institutes of Health R01-EY023966, R01-EY025011, and P30-EY008098, Eye and Ear Foundation (Pittsburgh, PA, USA), and Research to Prevent Blindness (support to UPMC Department of Ophthalmology).
Disclosure: A. Gogola, None; N.-J. Jan, None; K.L. Lathrop, None; I.A. Sigal, None